Hospital healthcare utilizers with medical and psychiatric claims in the Netherlands: a nationwide study

The Dutch Healthcare Authority (‘Nederlandse Zorgautoriteit’) provided claims data for psychiatric and medical (both inpatient and outpatient) hospital care in 2010–2011, which comprised the total number of claims for Dutch specialist care in these years. In the Netherlands, separate funding systems are in place to reimburse both types of care and comprise a case-mix system using a ‘Diagnosis Related Groups’ system, called ‘DBCs’ [20]. The claims data were anonymized, and this study did not intervene in human subjects. Hence, Dutch law did not require the approval of the Medical Research Ethics Committee and we followed all the necessary guidelines laid by Dutch law (7:457 lid 1 BW, https://​english.​ccmo.​nl/​investigators/​legal-framework-for-medical-scientific-research/​your-research-is-it-subject-to-the-wmo-or-not).

In 2010, the Netherlands had 16.6 million inhabitants with mandatory health insurance. Records from psychiatric and medical hospital care were linked, based on anonymized social security numbers. To ensure reliable linkage, birth year, sex, and residential district codes were also accessed. Figure 1 depicts the selection of patients for whom treatment began or ended in 2010. Patients below 18 and above 75 years of age were excluded because in the Netherlands psychiatric care is financed separately for patients < 18 years, and for the elderly, if cared for in long-term care facilities. Also excluded were psychiatric cases without diagnoses and non-fundable care for ‘adjustment disorders’, and ‘other problems that may be a focus of clinical attention’. The resulting database comprised 5,133,295 unique patients (n = 4,532,250 in medical care; n = 601,045 in psychiatric care). These were inpatients and outpatients, treated in Dutch specialist care, comprising all general and mental health hospitals and outpatient clinics.

Both psychiatric and general medical diagnoses were obtained directly from DBC data. Medical DBCs employed a description of the medical problem for coding based on the International Classification of Disease (ICD-10 [21]), while psychiatric DBCs employed the Diagnostic Statistical Manual of Mental Disorders-IV (DSM-IV-TR) classification [22]. In order to use one classification, the standardized and validated Dutch Healthcare Authority translation table was employed to convert all the DBC data according to the ICD-10. Individual healthcare utilization was assessed by counting psychiatric and medical diagnoses in the study period. If an individual showed more than one (psychiatric or medical) DBC with the same diagnosis, the diagnosis was deemed to be present only once. Then healthcare utilizers who only used medical healthcare were classified as ‘healthcare utilizers with medical claims’ (HUMC). Similarly, healthcare utilizers who only used psychiatric healthcare were classified as ‘healthcare utilizers with psychiatric claims’ (HUPC). Finally, healthcare utilizers with co-occurring medical and psychiatric healthcare utilization were classified as ‘healthcare utilizers with medical and psychiatric claims’ (HUMPC).

The Proportional Occurrence Difference (POD) and Proportional Occurrence Ratio (POR) were employed to express healthcare utilization differences between the groups just specified. The POD and POR were calculated using the following formulae:

A POR of > 1 indicated the higher proportional occurrence or over-representation of a psychiatric sub-chapter in HUMPCs compared with its occurrence in HUPCs, while a POR of < 1 indicated the under-representation of the specific subchapter.

In health services research, the standardization of rates is often based on direct or indirect arithmetic methods. When individual data are available, logistic regression models can be used to calculate adjusted rates [23]. The dependent variable in the logistic regression analysis was comorbidity (1 = present; 0 = absent), with age (continuous), sex, and the interaction of age and sex as independent variables. Sensitivity analyses were performed for some important diagnostic groups, since this resulted in minor differences only crude values are reported.

Table 2 shows psychiatric healthcare use broken down for men and women. ‘Mood [affective] disorders’ (27.8%) and ‘neurotic, stress-related and somatoform disorders’ (20.4%) were the most frequent diagnoses in psychiatric care. Women were more often than men diagnosed with ‘behavioral syndromes associated with physiological disturbances and physical factors’ (including eating disorders), ‘neurotic, stress-related and somatoform disorders’, mood [affective] disorders’, and ‘personality disorders’. Men were more often diagnosed with ‘disorders of psychological development’ including autism, ‘substance abuse disorders’, ‘schizophrenia, schizotypal and delusional disorders’ and ‘behavioral and emotional disorders with onset usually occurring in childhood and adolescence’ including attention deficit hyperactivity disorder.

In 2010, 217,887 (36.3%) of psychiatric healthcare utilizers also used medical care. Of all medical healthcare utilizers, 4.8% also used psychiatric healthcare. The following psychiatric disorder groups were found more often in HUMPCs compared to HUPCs: ‘organic mental disorders’ (POR 1.82), ‘behavioral syndromes associated with physiological disturbances and physical factors’ (POR 1.26), ‘mood [affective] disorders’ (POR 1.10), ‘neurotic, stress-related and somatoform disorders’ (POR 1.09), and ‘disorders of adult personality and behavior’ (POR 1.07). In contrast, ‘disorders of psychological development’ (POR 0.56) ‘schizophrenia, schizotypal and delusional disorders’ (POR 0.73), ‘behavioral and emotional disorders with usual onset occurring in childhood’ (POR 0.75), and ‘mental and behavioral disorders due to psychoactive substance abuse’ (POR 0.85) were found less frequently in HUMPCs compared to HUPCs (Table 3).

Of all psychiatric claims in 2010, HUPCs (63.7%) claimed a total of €2.75 billion and HUMPCs (36.3%) accounted for €1.78 billion. Thus, HUMPCs claimed 65% of total psychiatric costs. Median psychiatric claims of HUMPCs were 10% higher overall compared to HUPCs. When medical healthcare was utilized, median psychiatric claims were higher in four sub-chapters: + 26% in substance use disorders, + 17% in disorders of psychological development, + 12% in mood disorders, and + 1% in disorders with onset usually occurring in childhood. In three sub-chapters, psychiatric claims were not affected by medical healthcare utilization (disorders associated with psychological disorders and physical factors, organic disorders, and neurotic, stress-related, and somatoform disorders). Finally, in two sub-chapters, psychiatric claims were lower in HUMPCs than HUPCs: − 22% in personality disorders and − 8% in schizophrenia spectrum disorders, respectively (see Table 4).

In 2010, HUMCs claimed €9.59 billion. With respect to HUMPCs, €1.12 billion was claimed for medical costs, 10.5% of total medical claims. The median medical claims of HUMPCs were 40% higher than for HUMCs. A total of 15 medical diagnostic groups showed higher medical claims with co-occurring psychiatric care utilization. The most marked differences in patients with co-occurring psychiatric disorders (see Table 5) were observed in the diagnostic groups ‘endocrine, nutritional and metabolic diseases’ (94%), ‘congenital malformations, deformation and chromosomal abnormalities’ (89%), and ‘diseases of the ear and mastoid process’ (86%).

Medical healthcare utilization differences for psychiatric patients might be related to patient and health system factors [13]. In this subsection we concentrate on our findings with respect to these two (sets of) factors.

Inadequate healthcare utilization by psychiatric patients may result in higher claims expenditures of HUMPCs, as shown in our study. These patients have a higher risk for medical illness and experience more extended lengths of hospital stay. Restrictions in accessibility (as described above) and delayed diagnosis and treatment might lead to avoidable ED visits and (re) admissions [5]. LOS and readmissions were considered important cost-drivers of hospital care [36] and might thus explain many of the excess claims that we found for HUMPs compared to HUPCs and HUMCs.

In contrast, diagnoses of personality disorders and psychotic spectrum disorders were associated with lower psychiatric claims in HUMPCs than HUPCs. Personality disorders are known to pose a high economic burden [37]. Thus, it is remarkable that we found lower median costs of psychiatric claims for HUMPCs than HUPCs in these patient groups. A possible explanation is the lower probability of admission to specialized care for patients with personality disorders found by van Veen et al. [38]. A possible reason for the lower number of psychiatric claims for psychotic spectrum disorders in HUMPCs compared to HUPCs was that these patients tended to avoid healthcare, presenting first in medical hospitals and only when they had developed acute medical symptoms, as found by Swildens et al. [17]. Future research should examine how care can be effectively improved to alleviate this situation.

Our study found that HUMPCs, in almost all subgroups, had increased healthcare costs compared to HUMCs and HUPCs. Because patients with a mental disorder are more likely to have a physical disorder, it is important that physical disorders are detected early. The efforts to improve physical health in mentally ill patients could be to focused on patients with severe mental disorders [39]. Our study shows that a focus on public health is important for all patients with mental health problems. This should take into account, social determinants – the conditions in which people are born, grow up, live, work and age – and unhealthy lifestyles are intertwined and negatively affect the risk for physical and psychiatric disorders [13]. It is important to intervene early in life on social and lifestyle risk factors that significantly reduce the risk of physical and psychiatric illness disorders [13]. In the long run, hospital healthcare utilization may decrease due to these efforts.

In the Netherlands health care insurance is mandatory, thus replication of our results will not easily be accomplished in countries with different accessibility of care. On the other hand, the results will most likely be replicable for citizens with an inclusive health insurance package. A further limitation may be that this nationwide data was 10 years old, but since no major reforms have taken place, it can be assumed that general healthcare utilization patterns were likely to have been stable over time. It is reassuring that our findings are in line with those reported in recent literature. Hence, we consider these results to be representative both currently and for other countries with inclusive health insurance packages. A significant finding was that median claims in specialist medical care were considerably lower than median claims in psychiatric hospital care. This can be explained by a greater emphasis on outpatient care within specialist medical care. Many outpatients’ DBCs have low reimbursement rates, and thus median claims for medical care are lower than for psychiatric hospital care. In this nationwide study, we used POD and POR. Other studies that examined the impact of concurrent medical and psychiatric disorders on health utilization used odds ratios (OR) [9‐12] or latent class analyses [27, 28]. The calculation of the POR is comparable to that of the OR, and our results are therefore comparable to other studies. We only carefully speculated about possible patient and health system mechanisms that may explain the differences in healthcare utilization and claims in this study, however, our data were insufficient to formally examine these interpretations.