Success Factors for Effective and Efficient Prevention of Mental
Disorders in Dutch General Practice
Tosca Vennemann
1Supervisor: Prof. Dr. R..J.M. Alessie University of Groningen, The Netherlands
January 8th 2020
Abstract
This study examines to what extent factors related to mental health nurse practitioners (MHNPs), i.e. characteristics of the professional MHNP, the function MHNP and the care provided, are associated with the short-term (three months) effectiveness and efficiency of the care provided by the MHNP to adult patients with depressive and/or anxiety symptoms in Dutch general practice. Effectiveness is measured in terms of improvements in quality of life, and efficiency in terms of a positive net monetary benefit (NMB). The study concerns a prospective cohort study of adult patients who consulted MHNPs. The analyses are based on baseline measurement and short-term follow-up, which was three months after the intake by the MHNP. Results showed that patients gained on average 0.022 QALYs in three months. The mean total costs per patient were €5,211 (95% CI, €4,550 - €5,940) at intake and €4,889 (95% CI, €3,685 - €6,137) three months thereafter. MHNPs with working experience as an MHNP of more than eight years and mindfulness and/or relaxation exercises were associated with negative changes in the quality of life, but the more experienced MHNPs were shown to treat more severe patients. Additionally, older MHNPs and MHNPs working on a secondment basis rather than being employed by a general practice or self-employed were associated with positive NMB. Results showed that, during the three months follow-up, patients’ quality of life improved and healthcare costs decreased compared to baseline, which may indicate a positive effect of treatment by MHNP.
Keywords: Mental Health Nurse Practitioner, Mental Disorders, Depression, Anxiety, Utility, Costs,
Net-Benefit, Netherlands
JEL classification: I11, I19, I30
2
Preface
3
1. Introduction
Depression and anxiety are common mental disorders in all age groups. Approximately one out of five Dutch adults will experience a depressive (20.2%) or anxiety (19.6%) disorder at some point in their lives (de Graaf, ten Have, & van Dorsselaer, 2010). No clear improvement in mental health status has occurred in about a decade (de Graaf, ten Have, van Gool, & van Dorsselaer, 2012), whereas the consequences of these disorders in terms of lost health are huge. Depression is ranked by the World Health Organisation (WHO) as the single largest contributor to global disability (7.5% of all years lived with disability in 2015); anxiety disorders are ranked sixth (WHO, 2017). In addition to their health impact, these common mental disorders cause a significant economic burden due to lost economic output and their comorbidity with costly medical conditions, including cancer, cardiovascular disease, diabetes and obesity. Consequently, the prevention of depressive and anxiety disorders is recognized as a major public health challenge (Mnookin, Kleinman, & Evans, 2016).
Strengthening the provision of mental health care in the primary care setting is considered essential for achieving greater effectiveness and efficiency of prevention of common mental disorders. According to the WHO (2008), integrating mental health services into primary care is the most viable way of ensuring that people have access to the mental health care they need. This will enable people to access mental health services closer to their homes and indirect costs associated with seeking more specialized mental health care are avoided. Furthermore, delivering mental health services in primary care minimizes stigma and discrimination. Nonetheless, primary care systems must be strengthened before mental health integration can be reasonably expected to flourish.
In 2014, the Dutch government introduced reforms of the mental health care system with the aim to substitute secondary mental health care by primary mental health care. The intention was to realise better cost control of mental health care, while at the same time improving the quality of care (Trend repport GGZ, 2014). As an important part of the major reform, the function of mental health nurse practitioner (MHNP) has been implemented on a nationwide scale in Dutch general practice. Together with the general practitioner (GP), the MHNP acts as the gatekeeper to more specialised mental health care facilities. The MHNP supports the GP in treating patients with common mental health problems, such as depressive and anxiety symptoms. The MHNP provides support, guidance and short-term treatment, and aims to prevent the development of full-blown common mental disorders that require referral to more specialised mental health care.
4 who consulted an MHNP increased significantly in recent years, from almost 427,000 in 2015 to 536,000 patients in 2018 (Vektis, 2017).
The mental health care reforms aimed to substitute secondary mental health care by primary mental health care. The question, however, is whether this indeed has occurred (Magnée et al., 2019). Current waiting times for mental health services do not seem to indicate this substitution, which could mean that with the reforms the treatment for patients is just delayed. This would be undesirable because mental health problems could exacerbate by not providing appropriate treatment in time.
Given the crucial role of the function MHNP in strengthening mental health care in Dutch general practice, it is important to gain insight into the effectiveness and efficiency of the care provided by the MHNP, and which factors influence effective and efficient care (Magnée, de Beurs, Schellevis, & Verhaak, 2018; van Es & Nicolaï, 2015). MHNP-related characteristics are a relevant category of influencing factors to examine. The function MHNP is relatively new and under development. Previous research has indicated that substantial regional and local variation exists in the organisation and execution of the function MHNP in daily practice (Trend repport GGZ, 2014). Currently, the MHNP is not a profession, meaning that different professions can fulfil the function (Landelijke Huisartsen Vereniging (LHV), 2014). Consequently, MHNPs have different educational backgrounds. The most common three educational backgrounds are those of psychology, social psychiatric nursing and social work. Moreover, some of the MHNPs have followed an MHNP post-bachelor’s training, while other did not. Other background characteristics that vary among MHNPs include gender, age and working experience, both as a mental health care professional and as an MHNP. Additionally, organizational characteristics of the function MHNP differ in daily practice, such as the collaboration between MHNP and GP, and type of employment arrangement. Some MHNPs are employed by a general practice, while others are self-employed or seconded through intermediary organisations such as a primary care cooperation or a mental health care organisation (van Es & Nicolaï, 2015). Previous research has also indicated variation in the care provided by MHNPs in terms of the number of consultations per patient and referral behaviour. The most extreme example was founded by a health centre where the main task of the MHNP was to refer patients to the most appropriate treatments as quickly as possible (Trend repport GGZ, 2014). The variation among the MHNPs might affect the effectiveness and the efficiency of the care provided by the MHNP. Some MHNP-related factors might result in more effective and efficient care (i.e. in terms of gains in quality of life or reductions in resource use) than others.
5 characteristics of the professional MHNP, the function MHNP and the care provided, are associated with the short-term (three months) effectiveness and efficiency of the care provided by the MHNP to adult patients with depressive and/or anxiety symptoms. Effectiveness is measured in terms of improvement in quality of life, and efficiency is assessed in terms of a positive net monetary benefit (NMB). Identifying influencing MHNP-related factors is relevant to further develop the function MHNP in order to promote effective mental health care in Dutch general practice against reasonable costs.
This paper is organized as follows. Section 2 describes the study design and study population. Section 3 discusses the measures of the study, consisting of the dependent variables, potential predictors and potential confounders. Section 4 discusses the methodology used to select predictors and confounders and the regression analyses. Section 5 presents the results. Section 6 discusses the results and some limitations of the study. The final section closes with some concluding remarks.
2. Study design and study population 2.1 Study design
Since MHNPs are working in the vast majority of Dutch general practices, it was not possible to use a study design that included a comparison with a control group of general practices without an MHNP. Instead, a natural evaluation study was conducted, evaluating variation among MHNPs-related factors to changes in quality of life of the patient and NMB. Before the NMB evaluation the costs which were used as a stepping stone for the NMB analysis will be discussed.
This empirical study has used data that has been collected as part of a research conducted by the Trimbos Institute. A prospective study was conducted in which adult patients with depressive and/or anxiety symptoms and who had an intake consultation with an MHNP, were followed-up for three months.
2.2 Study population
6 with the MHNP, had sufficient knowledge about the Dutch language and cognitive skills to fill in a questionnaire and had access to the internet and an email address.
Of the eligible patients who were invited by the MHNPs, 713 patients were interested to participate (Figure 1). These patients received study-related information, and 483 patients (67.8%) gave their consent to participate and completed self-report questionnaires within two weeks after the intake consultation with the MHNP (baseline measurement, T0). Of these participating patients, 376 patients (77.8%) completed the questionnaires again at three-month follow-up (T1). A non-response analysis regarding demographic and health-related characteristics was conducted to identify whether respondents differed from non-respondents. Only the fraction of males was slightly lower in the respondents’ group. Patients who were referred to (more) specialised mental health care facilities were excluded (n=44). Patients were considered to have received care from an MHNP when they had at least two consultations with an MHNP and when they had received psychological education (guidance and advice), short-term self-help/psychological guidance, relapse prevention or monitoring/long-term guidance during at least one consultation. Patients who did not meet this definition of MHNP care were excluded (n=11). Resulting in a study population of 321 patients.
Figure 1. Flow chart. *Patients who had at least two consultations with
7
3. Measures
The study relied on self-reported data by patients and MHNPs. Data was collected in the following three ways. First, both at T0 (within two weeks after intake consultation with the MHNP) and T1 (at three-month follow-up), patients completed self-report questionnaires on socioeconomic characteristics, depressive and anxiety symptoms, quality of life, healthcare consumption and productivity losses, respectively. Second, at the start of their study participation, MHNPs completed self-report questionnaires about their background characteristics and the organisation of the function MHNP in daily practice. Third, after every consultation with a participating patient, MHNPs recorded the characteristics of the care provided (see Appendix A for questionnaires).
3.1 Dependent variables
Change in quality of life (QALY)
According to the guidelines for economic evaluations in healthcare provided by Care Institute Netherlands (2015), the EuroQol questionnaire (EQ-5D-5L) has been used to calculate (changes in) health related quality of life in terms of Quality Adjusted Life Years (QALY) gained. The EQ-5D-5L consists of five dimensions: mobility, self-care, usual activities, pain-discomfort and anxiety/depression. Each dimension has 5 levels: no problems, slight problems, moderate problems, severe problems and extreme problems. The patient is asked to indicate the state of their health by ticking the box at the most appropriate statement in each of the five dimensions. The decision results in a one-digit number that expresses the level selected for that dimension. The digits for the five dimensions can be combined into a five-digit number that describes the patient’s health state (EuroQol Research Foundation, 2017). The Dutch tariff for the EQ-5D-5L established by Versteegh et al. (2016) has been used to compute utilities. This utility value set represents the views of the Dutch population about the EQ-5D-5L health states. The change in QALYs in the three months was calculated by the area-under-the-curve (AUC) method using the EQ-5D-5L health states at T0 and T1 (Appendix B). In general, the change in QALYs ranges between minus one and one. Since this study focussed on three months, values ranged between -0.25 and 0.25 given that this is the maximum change in QALYs possible in three months. A negative value indicated a deterioration of a patient’s quality of life at three-months follow-up, while a positive value indicated an improvement.
Net monetary benefits (NMB)
8 QALY (Brouwer, van Baal, van Exel, & Versteegh, 2019). The present study focussed on patients with depressive and/or anxiety symptoms who received care from an MHNP. This can be considered as preventive care aimed at preventing the development of full-blown depressive and anxiety disorders. Therefore, a gain of one QALY was set conservatively at €20,000.
Costs
The Treatment Inventory of Costs in Patients with psychiatric disorders (TIC-P) was used to measure the healthcare consumption and productivity losses in the past three months for participating patients at both T0 and T1. TIC-P is a fairly feasible and reliable instrument for collecting data on medical consumption and productivity losses in patients with mild to moderate mental health problems (Bouwmans et al., 2013). Moreover, information about the number and type of consultations with the MHNP was available for each patient, as MHNPs recorded the care they provided to participating patients.
9
3.2 Potential predictors and confounders
Given that a large number of characteristics were considered, a selection was made based on expert opinion and statistical selection procedures (Appendix D). An expert panel (consisting of two MHNPs, one GP, one representative of a patient organisation and three researchers including one professor) selected a priori characteristics that were essential from a policy perspective to be included in the regression analyses as potential predictors of change in QALYs and NMB. Characteristics were divided into characteristics of the professional MHNP and the function MHNP, characteristics of care provided by the MHNP and patients’ characteristics.
Potential predictors: characteristics of the professional MHNP and the function MHNP
The considered potential predictors regarding the characteristics of the professional MHNP and the function MHNP are presented in Table 1. From an expert point of view the following factors were selected a priori: having a nursing background; pursuing or having completed an MHNP post-bachelor’s training degree; years of work experience as an MHNP; having 10 or more consultations on an average 8-hour working day; and being employed on a secondment basis (irrespective of the type of intermediary organisation). Besides these characteristics, the other characteristics were considered by a statistical selection procedure (see section 4).
Table 1. Characteristics of the professional MHNP and the function MHNP
A priori Age
Gender
Nursing backgrounda √
MHNP post-bachelor’s training degree (in progress or completed) √
Work experience as a mental health care professional
Work experience as an MHNP √
Working hours per week as an MHNP
Employed on a secondment basis (reference: employed by a general practice or self-employed) √
More than 10 consultations on an average 8-hour working day √
Regular consultation moments with GP
Average waiting time for new patients less than 2 weeks
Using questionnaires to support problem clarification or triage (i.e. often/always or not/sometimes) Using questionnaires to monitor outcomes (i.e. often/always or not/sometimes)
Having received >20 hours of training (in the previous 12 months)b
Having received intervision or supervision >6 times (in previous 12 months)c
a Included: master psychology, bachelor of nursing, bachelor social work with mental health care differentiation, social
psychiatric nurse, mental health care professional nurse, higher professional (applied) psychology education, higher professional nurse education, post higher professional education cognitive behavioural therapist
b E.g. masterclasses, workshops, courses and conferences
10
Potential predictors: characteristics of care provided by the MHNP
The following characteristics of care as recorded by the MHNPs were considered as potential predictors: the number of consultations, the number of homework exercises, and the type(s) of treatment technique(s) provided during the consultation. The care provided was specified as follows: cognitive behavioural therapy techniques, behavioural activation, short-term treatment, mindfulness/relaxation exercises and solution oriented/problem solving treatment. None of these care characteristics were selected a priori; all were considered by the statistical selection procedure.
Potential confounders: patients characteristics
The following baseline (T0) characteristics of the patients were considered as potential confounders: age, gender, education, living situation, working situation, ethnic origin, severity of anxiety and depressive symptoms. The severity of anxiety and depressive symptoms were included to capture the impact on the change in quality of life and NMB. The widely used and validated questionnaires Generalized Anxiety Disorder (GAD-7) and Patient Health Questionnaire on Depression (PHQ-9) were used to measure the severity of the anxiety and depressive symptoms. These screeners are quick and user-friendly, improving the recognition rate of depression and anxiety and facilitating diagnoses and treatment (Spitzer, Kroenke, & Williams). GAD-7 and PHQ-9 scores were correlated with changes in quality of life and NMB. Although these factors were endogenous, they were not correlated with the other independent variables. A priori selected patients characteristics were age, gender, GAD-7 and PHQ-9 scores. These and the other characteristics were considered by the statistical selection procedure.
4. Methodology
4.1 Selection of potential predictors and confounders
11 used for building a model. This procedure was repeated iteratively five times to choose the best fitting model and corresponding shrinkage parameter.
4.2 Main analysis
Stata 12.0 has been used for the regression analyses. Given the relatively low number of missing values, missing data was imputed using median imputation for continuous variables and an ‘unknown’ category was added for categorical variables.
Change in quality of life (QALY)
A regression model has been used to examine the associations of selected MHNP-related characteristics, while controlling for selected confounding patient variables, with health improvements of patients. The following model was used:
Δ𝐻𝑖1= 𝛽0+ 𝛽1𝑀𝐶𝑗0′ + 𝛽2𝐶𝐶𝑖′+ 𝛽3𝑃𝐶𝑖0′ + 𝜀i𝑡 (1)
The dependent variable Δ𝐻𝑖1 indicates the change in the quality of life of patient i between time 0 and
1. More specifically, the change in the quality of life of patient i in the three months after the intake by the MHNP. 𝑀𝐶𝑗0 represents a vector of characteristics of the professional MHNP and the function
MHNP at baseline of MHNP j. 𝐶𝐶𝑖′ stands for a vector of characteristics of care provided to patient i by
the MHNP. Characteristics of patient i at baseline are denoted by the vector 𝑃𝐶𝑖0′. 𝜀i𝑡represents the
standard errors for patient i at time t. Standard errors are clustered at the MHNP level to control for correlation between the standard errors of patients treated by the same MHNP.
Costs
12
Net monetary benefit (NMB)
A NMB regression analysis has been used to identify MHNP-related characteristics associated with greater net benefits while controlling for selected confounding patient variables. The following model was used:
NMB𝑖1= 𝛽0+ 𝛽1𝑀𝐶𝑗0′ + 𝛽2𝐶𝐶𝑖′+ 𝛽3𝑃𝐶𝑖0′ + 𝜑𝑖𝑡 (2)
The NMB model differs relative to the change in quality of life model only in the dependent variable. The dependent variable is now NMB𝑖1, which indicates the net monetary benefit of patient i in three
months after intake by the MHNP. The vectors of independent variables are the same as in the change in quality of life model. 𝜑𝑖𝑡 represents the error term for patient i at time t. Standard errors clustered
at the MHNP level are computed to control for correlation between the error terms of patients treated by the same MHNP.
Sensitivity analyses
Two sensitivity analyses were conducted. First, the main analyses were repeated while excluding patients who reported to have received care from a mental health care professional in the three months before intake consultation with the MHNP (T0) (n= 94). Patients who reported a maximum of two consults (intake and first treatment) by a psychologist having an own practice were not excluded since the patient may have been confused in seeing the MHNP as a psychologist. Second, the main analyses were repeated while excluding patients with minimal anxiety and depressive symptoms at T0 (i.e. having a GAD-7 baseline score and a PHQ-9 baseline score between zero and four) (n= 38).
5. Results
5.1 Characteristics of the study population
13
Table 2. Baseline patient characteristics (at intake consultation)
Responders% (N=321)
Missing
Female gender 69.5
Mean age at intake (range) 41.9 (18-72) 1
18-30 26.2
30-45 33.0
45-59 26.8
>60 14.0
Living situation
Married or living with a partner 66.7 Other (single, divorced, widow) 33.3 Educational levela Low 8.4 Middle 52.3 High 39.3 Working situation 11 Employee or entrepreneur 74.2 Disabled or unemployed 16.8
Other (housewife/man, school, retirement) 9.0 Ethnicityb
Netherlands 86.3
Migration background 13.7
Mean severity anxiety symptomsc (range) 8.4 (0-20)
Minimal (0-4) 19.3
Mild (5-9) 42.7
Moderate (10-14) 27.1
Severe (15-21) 10.9
Mean severity depression symptomsd (range) 9.4 (0-27)
Minimal (0-4) 19.0
Mild (5-9) 36.5
Moderate (10-14) 26.5
Severe (15-27) 18.1
a Categorised based on Dutch educational system. Low: less than primary education,
primary education, lower secondary education. Middle: higher secondary education, lower vocational education. High: higher vocational education, university.
b Categorised based on CBS definition; migration background when patient was born
abroad or at least one parent was born abroad
c Based on GAD-7 d Based on PHQ-9
14 professional. An MHNP worked on average 4.8 years as an MHNP. MHNP worked on average 24.2 hours per week, with 15.6% working less than 16 hours, 45.3% worked 16-28 hours and 39.1% worked more than 28 hours and had on average 9.1 consults per workday. Most MHNPs worked on a secondment basis (65.6%) rather than employed by a general practice or being self-employed (34.4%). In general, considerable variation existed among the MHNPs regarding he measured characteristics. More details are shown in Table 3.
Table 3. Characteristics MHNP
MHNP% N=64
Female gender 68.8
Mean age (range) 51.2 (33-63)
<45 18.7
45-55 46.9
>55 34.4
Background and experience
Nursing background 73.4
MHNP post-bachelor’s training degree (in progress or completed) 59.4 Mean work experience as a mental health care professional in years (range) 22.3 (1-41) Mean working experience as an MHNP in years (range) 4.9 (0-14)
<2 18.8
2-4 29.7
4-8 35.9
≥8 15.6
Mean working hours per week as MHNP (range) 24.1 (6-40)
<16 15.6
16-28 45.3
≥28 39.1
Mean consultations on an average 8-hour working day (range) 9.1 (4-15)
<10 face to face consults 56.2
≥10 face to face consults 43.8
Employment
Employed by a general practice or self-employed 34.4
On a secondment basis 65.6
Regular consultation moments with GP 53.1
Average waiting time for new patients
<2 weeks 60.9
>2 weeks 39.1
15
5.2 Selected potential predictors and confounders
The selection a priori and LASSO method selected in total 13 MHNP-related predictors and patient-related confounders for the model predicting change in quality of life and 12 characteristics for the model predicting NMB (Appendix D). The a priori selected characteristics (k=9), as discussed in section 3, were part of both models. The LASSO method selected number of working hours per week as an MHNP, using mindfulness and/or relaxation exercises by the MHNP, patient’s educational level and patient’s work situation for the model predicting the change in quality of life. MHNP’s age, patient’s work situation and patient’s ethnicity were additionally selected by the LASSO method for the model predicting NMB. Since the number of consultations by the MHNP was not selected by the LASSO method for the quality of life model this factor has been included on the cost side. Resulting in a more complete picture of patient’s total costs.
Table 4. Descriptive statistics of dependent variables
Dependent variables N Mean Min Q1 Q2 Q3 Max
Δ𝐻𝑖1a 321 0.022 -0.102 0 0.019 0.039 0.219
NMB𝑖1b 294 -4443.99 -164197.60 -3574.47 -553.01 141.40 3230.69
Note: Descriptive statistics of study observations.
a Change in quality of life in three months.
b NMB in three months. The 27 missing observations resulted from the missing costs values.
5.3 Main Analysis
Change in quality of life
16
Table 5 Change in quality of life model
Δ𝐻𝑖1
Characteristics of the professional MHNP and the function MHNP
Nursing background 0.005
(0.004)
MHNP post-bachelor’s training degree 0.006
(0.004) Work experience as an MHNP <2 years Reference 2-4 years -0.004 (0.004) 4-8 years -0.003 (0.005) ≥8 years -0.014* (0.008) Working hours per week as an MHNP
<16 hours Reference
16-28 hours -0.009
(0.008)
≥28 hours 0.002
(0.008) ≥10 face to face consults on an average 8-hour workday 0.001
(0.004)
Employed on a secondment basis 0.004
(0.005) Characteristics of care provided by MHNP
Other treatment Reference
Mindfulness and/or relaxation exercises -0.008*
(0.004) Unknowna -0.009 (0.007) Characteristics of patient Age -0.000 (0.000) Female -0.000 (0.005) Educational level Low Reference Middle 0.014 (0.010) High 0.006 (0.010) Working situation
Employee or entrepreneur Reference
School, housewife/man or retired -0.007
(0.005)
Disabled or unemployed -0.004
17
Unknownb 0.021
(0.014)
Severity anxiety (GAD-7) 0.002**
(0.001) Severity depression (PHQ-9) 0.001* (0.001) Constant -0.013 (0.015) N 321 N-clusters 64 R-squared 0.196
Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses, clustered at
the MHNP level
a 27 missing observations b 11 missing observations
Costs
The mean total costs per patient in the past three months were €5,211 (CI, €4,550-€5,940) in the three-month period before intake consultation with the MHNP (T0), and €4,889 (CI, €3,685-€6,137) in the three-month follow-up period (T1) (Table 6), resulting in a mean decrease of €320 per patient. The healthcare costs were higher in the three-month follow-up period (€458) than the three-month period before T0 (€413). Note the positive psychologist cost. Psychologist costs were expected to be zero since eligible patients did not receive mental health care in the past. Patients could have been confused by seeing the MHNP as a psychologist. Consequently, MHNP consultation costs were counted twice, which means that the mean total costs per patients decreased further than €320 per patient. The lower mean total costs per patients were mainly explained by the changes in productivity and patient and family costs. Mean productivity costs were €4,131 (CI, €3,537-€4,781) at intake and €3,987 (CI, €2,930-€5,160) after three months. While absenteeism costs increased by €463, presenteeism costs decreased considerably by €606. Mean patient and family costs were €643 (CI, €487-€814) at intake and decreased in three months to €380 (CI, €291-€491), mainly due to the €284 decrease in unpaid work costs.
Table 6 Mean costs per patient last 3 months (€, 2018)
Total costs (€) at intake (T0)
18
Psychologist 47.3 77.1
Outpatient clinic specialist 42.2 38.1
Addiction consultant 0.0 0.3
Occupational therapist 20.5 46.8
Total clinical consultations [95% CI] 346.7 [309.4-384.0] 346.9 [304.8-391.4] MHNP consultationa Face-to-face consultation Intake 17.4
Short consult (<20 min.) 3.0
Long consult (>20 min.) 5.9
Home visit (<20min.) 4.5
Home visit (>20min.) 7.4
Other (call, mail, feedback, group) 1.5
Total MHNP costs [95% CI] 39.7
[36.2-43.7] Other healthcare
Day/night treatment hospital or mental
health care centre 25.0 30.7
Psychotropic medication 41.3 41.0
Total other healthcare costs [95% CI] 66.3 [44.9-96.7]
71.7 [41.0-122.0]
Total healthcare costs [95% CI] 413.0
[367.5-464.4] 457.9 [402.1-537.0] Productivity costs Absenteeism 3108.4 3571.1 Presenteeism 1022.2 415.9
Total productivity costs [95% CI] 4130.6
[3536.8-4781.0]
3987.0 [2930.3-5159.8]
Patient and family costs
Unpaid work 620.2 336.0
Transport costs 22.4 43.8
Total patient and family costs [95% CI] 642.6 [487.4-814.21]
379.8 [290.5-490.5]
Total costs per patient [95% CI] 5211.2
[4550.0-5939.5]
4889.0 [3685.2-6136.5]
Note: Separate items represents the mean of the study observations whereas the total cost items represent the mean and 95% confidence interval after bootstrapping
a27 missing observations
NMB
19 significantly associated with a positive NMB of the care provided. Older MHNPs were associated with a positive NMB, i.e. more efficient care. However, working experience as an MHNP was not significant. As a result, MHNPs were working not necessarily more efficiently when they had more working experience as an MHNP. Furthermore, MHNPs who were working on a secondment basis provided significantly more efficient care than MHNPs who were employed by a general practice or self-employed. Four of the selected patient-related confounders were significantly related to NMB. Older patients were associated with a negative NMB, i.e. less efficient care. Moreover, the care provided to patients who were not an employee or entrepreneur was significantly more efficient than the care offered to employees and entrepreneurs. Finally, severity of depressive symptoms was negatively associated with NMB, whereas severity of anxiety symptoms was positively associated with NMB. Meaning that care provided to patients with anxiety complaints was efficient while care provided to patients with depressive complaints was inefficient. Whether the patient was male or female or had a migration background did not affect the NMB.
5.4 Sensitivity analyses
The results of the two sets of sensitivity analyses are shown in Appendix E. In the first sensitivity analyses, patients who reported to have received care from a mental health care professional in three months before intake consultation with the MHNP (T0) were excluded (n=94). Results showed that MHNP’s work experience and providing mindfulness and/or relaxation exercises were no longer significantly associated with change in quality of life. The severity of anxiety and depressive symptoms remained associated with a positive change in quality of life. Regarding the model predicting NMB, MHNP’s age was no longer significantly associated with NMB, while the positive effect of working on a secondment basis was robust. Compared to the original analyses using the total study population, a new finding was that the care provided by MHNPs with a nursing background was less efficient than the care offered by MHNPs with another professional background. Of the patient-related confounders, only the effect of patient’s work situation was largely robust.
20
Table 7 NMB model
NMB Characteristics of professional MHNP and function MHNP
Age 232.23* (119.51) Nursing background -15.23 (2,030.35) MHNP post-bachelor’s training 1,441.00 (1,856.76) Work experience as an MHNP <2 years Reference 2-4 years -1,831.18 (2,933.85) 4-8 years -1,899.81 (2,410.93) ≥8 years -2,657.57 (2,092.17) ≥10 face to face consults on an average 8-hour workday -1,243.99 (1,394.49)
Employed on a secondment basis 4,012.32**
(1,894.60) Characteristics of patient Age -148.69** (62.08) Female 1,102.52 (1,971.89) Working situation
Employee or entrepreneur Reference
School, housewife/man or retired 3,710.18***
(1,114.50) Disabled or unemployed 4,681.15*** (1,000.27) Unknowna 3,582.30* (1,985.38) Migration background -5,900.61 (4,093.18)
Severity anxiety (GAD-7) 276.20*
(144.96) Severity depression (PHQ-9) -454.59*** (159.60) Constant -9,958.85 (6,681.86) N 321 N-clusters 64 R-squared 0.141
Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses,
clustered at the MHNP level
21
6. Discussion
The study explored whether MHNP-related characteristics influence the short-term (three months) effectiveness and efficiency of the care provided by MHNPs to adult patients with depressive and/or anxiety symptoms in Dutch general practices.
Two MHNP-related characteristics were found to be negatively related to changes in quality of life, i.e. less effective care. First, patients of MHNPs with more than eight years of work experience as an MHNP showed less improvement in quality of life compared to patients of MHNPs who have worked less than two years in this position. This may be explained by the logic that the more severe patients may be perceived as challenging by more experienced MHNPs. Leading to the decision to treat these patients rather than referring them to (more) specialised mental health care. A post-hoc analyses looking into this matter revealed that MHNPs with more experience indeed treated relatively more patients with a lower quality of life compared to MHNPs with less than two years of work experience (i.e. baseline utility of 0.71 for the patient of MHNPs with more than eight years of experience compared to an average of 0.74 for the total study population, and 0.77 for the patients of MHNPs with less than two years of work experience). This indicates that less experienced MHNPs may be referring patients sooner to specialised mental health care. Second, mindfulness and/or relaxation exercises were negatively related to changes in quality of life, indicating that other treatments were more effective in improving patients’ health status. None of the other MHNP-related characteristics were found to significantly influence change in quality of life at three-month follow-up. Of the patient-related characteristics that were considered in the analysis, having more anxiety and/or depressive symptoms was associated with higher positive changes in quality of life. A possible explanation for this result is that patients in a more severe mental health state are more likely to improve by treatments provided by an MHNP in their quality of life than patients who are experiencing less symptoms.
22 symptoms the costs were not outweighed by the benefits whereas for patients with anxiety symptoms these were. This result indicate that the costs associated with depressive symptoms were higher than for anxiety symptoms.
For both sensitivity analyses, in general, it appeared that when the study population was limited to a more severe health state (i.e. removing patients with a low GAD-7 or PHQ-9 score), results were rather similar to the base case analyses. When excluding patients who reported to have visited a psychologist in the three months before intake consultation with the MHNP, it can be noticed that most of the predictors were no longer significant. This may be due to smaller improvements observed in the resulting population in terms of quality of life, and due to a smaller sample size (i.e. reduced power to detect statistical differences). Overall results appeared to be robust (especially for the NMB analyses).
Patients experienced a decrease in the mean total cost per patient within three months, which resulted largely from the increase in productivity (decrease in presenteeism and unpaid work costs). This result in combination with the on average positive QALYs seems to imply that seeing an MHNP improved the health of the patients.
Some explicit choices have been made in this study. For example, the number of consultations by the MHNP was included on the cost side since the LASSO method considered this factor to be redundant for the quality of life model. Nevertheless, the number of consultations by the MHNP could be a factor that impacts changes in the quality of life of a patient.
23 the sensitivity analyses. Psychologist costs were expected to be zero since eligible patients did not receive mental health care in the past. Although the formulation of the questionnaires tried to prevent this, patients could have been confused by seeing the MHNP as a psychologist. Meaning that the mean total costs per patients possibly decreased further than €320 per patient (since MHNP consultation costs were counted twice). However, the sensitivity analyses showed that the results were robust for this limitation. Third, the incomplete registration of MHNPs about the characteristics of the care provided to participating patients is a limitation. A more complete registration may identify other factors related to care provided by the MHNP. Fourth, the sensitivity of the EQ-5D-5L could be questioned since the questionnaire consists of just five questions. Information could have been lost by this short questionnaire. Meaning that identifying MHNP factors which were associated with effective and efficient care was difficult. Fifth, the WTP for one QALY was set relatively low (equal to the WTP for preventive interventions). However, the disease burden associated with early stages of depressive and anxiety disorders are possibly larger than in case of preventive medicine. The combination of a low WTP and relatively large costs at three months could have made it hard to draw conclusions about the cost effectiveness of certain factors. This is emphasized by the fact that most of the costs occurred in the three months follow-up phase. If patients would remain in the observed health state after a longer period of time, cost-effectiveness results may look more positive. Lastly, related to the previous argument, this study focussed on the short-term follow-up which was three months after the intake by the MHNP. It is important to study the long-term effects to see if improvements in QALYs and reductions in resource utilization are sustainable.
7. Conclusion
This study examined to what extent MHNP-related, factors, i.e. characteristics of the professional MHNP, the function MHNP and the care provided, were associated with effectiveness and efficiency of the care provided by the MHNP to adult patients with depressive and/or anxiety symptoms in Dutch general practice. Effectiveness was measured in terms of improvements in quality of life, and efficiency in terms of a positive NMB. The study concerned a prospective cohort study of adult patients who consulted MHNPs. This study focused on the short-term follow-up, which was three months after the intake by the MHNP.
24 MHNPs were shown to treat more severe patients. Additionally, older MHNPs and MHNPs working on a secondment basis rather than being employed by a general practice or self-employed were associated with positive NMB.
Because the WTP was set low, and the follow-up period was only three months, cost-effectiveness might improve with a longer follow-up period if the health state of patients stays stable or improves. Further research is needed to examine long-term effects and to examine factors associated with effective treatments by the MHNP in more detail. This may require a larger sample or a specific research design which allows for causal inference. Identifying these success factors may help to provide effective mental health care in general practice against reasonable costs.
25
References
Bouwmans, C., de Jong, K., Timman, R., Zijlstra-Vlasveld, M., van der Feltz-Cornelis, C., Tan, S. S., & Hakkaart-van Roijen, L. (2013). Feasibility, reliability and validity of a questionnaire on healthcare consumption and productivity loss in patients with a psychiatric disorder (TiC-P).
BMC health services research, 13(1), 217.
Brouwer, W., van Baal, P., van Exel, J., & Versteegh, M. (2019). When is it too expensive? Cost-effectiveness thresholds and health care decision-making. The European Journal of Health
Economics, 20(2), 175-180.
de Beurs, D., Prins, A., & Nielen, M. (2018). Psychische en sociale problematiek in de huisartsenpraktijk in de periode 2011–2017. Utrecht: NIVEL.
de Graaf, R., ten Have, M., & van Dorsselaer, S. (2010). De psychische gezondheid van de Nederlandse bevolking. Nemesis-2: Opzet en eerste resultaten, Trimbos-Instituut, Utrecht.
de Graaf, R., ten Have, M., van Gool, C., & van Dorsselaer, S. (2012). Prevalence of mental disorders and trends from 1996 to 2009. Results from the Netherlands Mental Health Survey and Incidence Study-2. Social psychiatry and psychiatric epidemiology, 47(2), 203-213.
EuroQol Research Foundation. (2017). EQ-5D-5L About. Retrieved from https://euroqol.org/eq-5d-instruments/eq-5d-5l-about/. Accessed on September 24th 2019.
Fonti, V., & Belitser, E. (2017). Feature selection using lasso. VU Amsterdam Research Paper in Business
Analytics.
Hakkaart-van Roijen, L., van der Linden, N., Bouwmans, C., Kanters, T., & Tan, S. S. (2015). Kostenhandleiding. Methodologie van kostenonderzoek en referentieprijzen voor economische
evaluaties in de gezondheidszorg. In opdracht van Zorginstituut Nederland. Geactualiseerde versie.
Independer. (2017). Huisartstarieven 2017. Retrieved from
https://www.independer.nl/huisarts/info/wat-kost-een-huisarts/huisartstarieven.aspx. Accessed on October 23th 2019.
Jiang, H., & Zhou, X. H. (2004). Bootstrap confidence intervals for medical costs with censored observations. Statistics in Medicine, 23(21), 3365-3376.
Landelijke Huisartsen Vereniging (LHV). (2014). Functie- en competentieprofiel POH-GGZ 2014.
Utrecht: Nederlands Huisartsen Genootschap (NHG).
Magnée, T., de Beurs, D., Schellevis, F., & Verhaak, P. (2018). Steeds belangrijkere rol huisartsenpraktijk bij psychische problemen. Huisarts en wetenschap, 61(11), 14-17.
26 Mnookin, S., Kleinman, A., & Evans, T. (2016). Out of the shadows: Making mental health a global
development priority. Washington DC: World Bank Group.
Spitzer, R. L., Kroenke, K., & Williams, J. B. Patient Health Questionnaire Screeners. Retrieved from https://www.phqscreeners.com/select-screener. Accessed on October 5th 2019.
Trend repport GGZ. (2014). Versterking van de GGZ in de huisartsenpraktijk. Terugblik, stand van zaken en vooruitblik. Utrecht: Timbos Instituut (Netherlands Institute of Mental Health and
Addiction).
van Es, M., & Nicolaï, L. (2015). Hoe gaat het met de praktijkondersteuner-ggz? Tijdschrift voor
praktijkondersteuning, 10(2), 42-48.
Vektis. (2017). Zorgthermometer Inzicht in de GGZ. Zeist: Vektis.
Versteegh, M. M., Vermeulen, K. M., Evers, S. M., de Wit, G. A., Prenger, R., & Stolk, E. A. (2016). Dutch tariff for the five-level version of EQ-5D. Value in health, 19(4), 343-352.
WHO. (2008). Integrating mental health into primary care: a global perspective. World Health
Organization.
WHO. (2017). Depression and other common mental disorders: global health estimates. World Health
Organization.
Zorginstituut Nederland. Farmacotherapeutisch Kompas Geneesmiddelen. Retrieved from https://www.farmacotherapeutischkompas.nl/. Accessed on October 10th 2019.
27
Appendices
Appendix A. Questionnaires
- Questionnaire about socioeconomic characteristics of the patient, EQ-5D-5L, GAD-7, PHQ-9, TIC-P (in Dutch)
- Questionnaire about the characteristics of the professional MHNP and the function MHNP (in Dutch)
- Questionnaire about the characteristics of the care provided by the MHNP (in Dutch)
Appendix B. Dutch tariff for the EQ-5D-5L
Table 8. Dutch tariff for the EQ-5D-5L (Versteegh et al, 2016)
Dimensions EQ-5D-5L Problems Level Mobility (=MO) Selfcare (=SC) Usual activities (=UA) Pain/discomfort (=PD) Anxiety/depressive (=AD) No 1 0 0 0 0 0 Slight 2 -0.035 -0.038 -0.039 -0.066 -0.070 Moderate 3 -0.057 -0.061 -0.087 -0.092 -0.145 Severe 4 -0.166 -0.168 -0.192 -0.360 -0.356 Extreme 5 -0.203 -0.168 -0.192 -0.415 -0.421
Note: The values represent the preferences of the Dutch population. More extreme problems result in a higher decrease in utility. The utility level is determined as described in model 3 (see below); if no problems on each dimension the utility level is 1, if any problem the constant is equal to 0.953 from which the negative utility value is added. The change in quality of life in three months is determined as described in model 4 (see below)
𝐻𝑖𝑡 = { 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 + 𝑀𝑂1 𝑖𝑓 𝑛𝑜 𝑝𝑟𝑜𝑏𝑙𝑒𝑚𝑠 (ℎ𝑒𝑎𝑙𝑡ℎ 𝑠𝑡𝑎𝑡𝑒 11111)
𝑖𝑡+ 𝑆𝐶𝑖𝑡+ 𝑈𝐴𝑖𝑡+ 𝑃𝐷𝑖𝑡+ 𝐴𝐷𝑖𝑡 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (3)
Δ𝐻𝑖1= (𝐻𝑖1− 𝐻𝑖0)/4 (4)
Appendix C. Reference prices
Table 9. Reference prices
Reference price Financial year Costs 2018
Clinical consultationa
General practitioner € 66.00 2015 € 68.27
Social worker € 65.00 2015 € 67.24
Physiotherapist € 33.00 2015 € 34.14
Psychologist mental health care centre € 98.00 2015 € 101.37 Psychologist own practice € 94.00 2015 € 97.23
Psychologist hospital € 79.00 2015 € 81.72
Outpatient clinic specialist € 91.00 2015 € 94.13
Addiction consultant € 80.00 2015 € 82.49
28
MHNP consultationb
Intake € 18.47 2017 € 18.79
Short consult (<20 min.) € 9.23 2017 € 9.39 Long consult (>20 min.) € 18.47 2017 € 18.79 Home visit (<20min.) € 13.85 2017 € 14.09 Home visit (>20min.) € 23.09 2017 € 23.49
Phone consult € 4.62 2017 € 4.70
Mail consult € 4.62 2017 € 4.70
Feedback consult € 4.62 2017 € 4.70
Group consult € 9.23 2017 € 9.39
Other healthcare costsa
Day treatment mental health care centre € 302.00 2015 € 312.39 Treatment mental health care centre including
overnight stay € 476.00 2015 € 492.37
Treatment hospital including overnight stay € 476.00 2015 € 492.37
Other costsa
Productivity costs per hour paid worker € 34.75 2015 € 35.95 Unpaid work, replacement costs per hour € 14.00 2015 € 14.48 Transportation costs per kilometres € 0.19 2015 € 0.20
Parking costs € 3.00 2015 € 3.10
Note: CBS consumers price index is used to convert prices to the price year 2018
aSource: Dutch manual for economic evaluation in healthcare (2015) bSource: Independer General Practitioner rates (2017).
Table 10. Transportation costs
Destination of care Average distance (kilometres) Return distance (kilometres) Costs kilometres 2018 General practitioner 1.1 2.2 € 0.43 Social worker 1.1 2.2 € 0.43 Physiotherapist 2.2 4.4 € 0.86
Psychologist own practice 1.1 2.2 € 0.43
Addiction consultant 3.7 7.4 € 1.45
Occupational therapist 1.1 2.2 € 0.43
Hospital 7 14 € 2.75
Mental health care centre 3.7 7.4 € 1.45
Pharmacy 1.3 2.6 € 0.51
Source: Dutch manual for economic evaluation in healthcare (2015)
Table 11. Psychotropic medication costs
Psychotropic medication Dosea
Recommended
dose per daya Mean costsb Costs per day
Anafranil (Clomipramine) 25mg 3 € 4.08 € 12.23
Cipramil (Citalopram) 40mg 1 € 1.08 € 1.08
Dalmadorm (Flurazepam) 30mg 1 € 2.93 € 2.93
29 Efexor (Venlafaxine) 225mg 1 € 6.51 € 6.51 Imovane (Zopiclon) 7,5mg 1 € 1.89 € 1.89 Lexapro (Escitalopram) 10mg 1 € 1.95 € 1.95 Mogadon (Nitrazepam) 5mg 1 € 1.33 € 1.33 Normison (Temazepam) 20mg 1 € 1.68 € 1.68 Nortilen (Nortriptyline) 25mg 3 € 5.68 € 17.04 Prozac (Fluoxetine) 20mg 1 € 3.06 € 3.06 Remeron (Mirtazapine) 30mg 1 € 1.53 € 1.53 Seresta (Oxazepam) 50mg 1 € 1.37 € 1.37 Seroxat (Paroxetine) 20mg 1 € 1.86 € 1.86 Stesolid (Diazepam) 10mg 1 € 0.75 € 0.75 Stilnoct (Zolpidem) 10mg 1 € 1.83 € 1.83 Temesta (Lorazepam) 1mg 2 € 3.57 € 7.13 Tryptizol (Amitriptyline) 25mg 2 € 2.20 € 4.40 Xanax (Alprazolam) 1mg 3 € 3.96 € 11.87 Zoloft (Sertraline) 50mg 1 € 2.25 € 2.25 aSource: www.farmacotherapeutischkompas.nl bSource: www.medicijnkosten.nl
Appendix D. Selection potential predictors and confounders
Table 12. Selection potential predictors and confounders
A priori LASSO FINAL
U&NMB U NMB U NMB
Potential predictors: characteristics of professional MHNP and function MHNP
Age √ √
Gender
Nursing backgrounda √ √ √
MHNP post-bachelor’s training degree (in progress or
completed)
√ √ √
Work experience as a mental health care professional
Work experience as an MHNP √ √ √
Working hours per week as a MHNP √ √ Employment (employed by general practice/self-employed or on secondment basis) √ √ √ √
Number of consults per workday √ √ √
Regular consultation moments with GP
Average waiting time for new patients is <2 weeks
30 Questionnaires to monitor
outcomes
Having received >20 hours of training (in the previous 12 months) (e.g. masterclasses, workshops, courses and conferences)
Having received intervision or supervision >6 timesb (in previous 12 months)
Potential predictors: characteristics of care provided by MHNP
Number of contacts with MHNP N/A N/A
Number of homework exercises Relapse prevention
Cognitive behavioural therapy Behavioural activation Short-term treatment
Mindfulness/relaxation exercises √ √
Solution oriented/problem solving
Potential confounders: characteristics of patient
Age √ √ √ √ Gender √ √ √ Education √ √ Living situation Working situation √ √ √ √ Ethnicity √ √
Severity anxiety (GAD-7) √ √ √ √ √
Severity depression (PHQ-9) √ √ √ √
K=9 K=6 K=6 K=13 K=12
aIncluded: master psychology, bachelor of nursing, bachelor social work with mental health care
differentiation, social psychiatric nurse, mental health care professional nurse, higher professional (applied) psychology education, higher professional nurse education,
post higher professional education cognitive behavioural therapist
31
Appendix E. Sensitivity analyses
Table 13. Sensitivity analyses of changes in quality of life model
Exclude psychologist consults Excluding minimal complaints Δ𝐻𝑖1 Δ𝐻𝑖1
Characteristics of professional MHNP and function MHNP
Nursing background 0.001 0.005
(0.005) (0.005)
MHNP post-bachelor’s training degree -0.001 0.006
(0.006) (0.005)
Work experience as an MHNP
<2 years Reference Reference
2-4 years -0.006 -0.001 (0.006) (0.005) 4-8 years -0.005 -0.003 (0.008) (0.006) ≥8 years -0.002 -0.014 (0.010) (0.009)
Working hours per week as an MHNP
<16 hours Reference Reference
16-28 hours 0.003 -0.010
(0.011) (0.009)
≥28 hours 0.009 0.000
(0.011) (0.009)
≥10 face to face consults on an average 8-hour workday
0.005 0.001
(0.006) (0.005)
Employed on a secondment basis 0.006 0.006
(0.007) (0.005)
Characteristics of care provided by MHNP
Other treatment Reference Reference
Mindfulness and/or relaxation exercises -0.005 -0.010**
(0.006) (0.004) Unknowna -0.011 -0.010 (0.008) (0.009) Characteristics of patient Age 0.000 -0.000 (0.000) (0.000) Female -0.002 0.002 (0.006) (0.005) Educational level
Low Reference Reference
Middle 0.020 0.018
(0.012) (0.013)
High 0.019 0.008
32 Working situation
Employee or entrepreneur
School, housewife/man or retired -0.003 -0.010
(0.006) (0.007)
Disabled or unemployed -0.006 -0.000
(0.009) (0.007)
Unknownb 0.004 0.023
(0.009) (0.015)
Severity anxiety (GAD-7) 0.002** 0.002**
(0.001) (0.001) Severity depression (PHQ-9) 0.001** 0.001* (0.001) (0.001) Constant -0.034* -0.016 (0.019) (0.017) N 227 283 N-clusters 62 61 R-squared 0.175 0.186
Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses, clustered at the MHNP
level
a 27 missing observations b 11 missing observations
Table 14. Sensitivity analyses of NMB
Exclude psychologist consults Excluding minimal complaints NMB NMB
Characteristics of professional MHNP and function MHNP Age 48.60 230.13* (74.06) (126.45) Nursing background -2,305.36** -69.13 (1,030.93) (2,205.95) MHNP post-bachelor’s training 1,115.18 1,619.26 (1,289.80) (2,084.76) Work experience as an MHNP <2 years 2-4 years -679.67 -1,526.16 (1,617.95) (3,230.16) 4-8 years -2,965.69 -1,290.39 (1,812.64) (2,672.76) ≥8 years -3,410.60 -2,053.39 (3,133.89) (2,269.97) ≥10 face to face consults on an average 8-hour
workday
-1,059.14 -1,353.12 (1,033.54) (1,538.87)
Employed on a secondment basis 4,264.27*** 4,279.74**
33 Characteristics of patient Age -30.16 -166.63** (57.45) (69.05) Female 747.56 1,157.25 (1,335.55) (2,249.89) Working situation Employee or entrepreneur
School, housewife/man or retired 3,902.79*** 4,130.87*** (1,030.25) (1,336.88) Disabled or unemployed 2,356.58 5,017.45*** (1,731.89) (1,029.47) Unknowna 5,396.38** 3,964.76* (2,503.61) (2,173.98) Migration background -3,596.87 -6,136.54 (2,621.00) (4,314.58)
Severity anxiety (GAD-7) 90.96 327.88**
(179.30) (159.95) Severity depression (PHQ-9) -206.31 -438.59** (188.96) (165.22) Constant 23,991.85*** -10,611.02 (5,496.99) (7,005.38) N 227 283 N-clusters 62 63 R-squared 0.13 0.14
Note: *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors in parentheses, clustered at the MHNP
level