Small area analysis is a health services research technique that facilitates geographic comparison of health services utilization rates [1
]. Using this technique, researchers have consistently documented the existence of supplier-induced demand [2
] for health services [3
]. Rates of procedures--such as tonsillectomy, prostatectomy, and hysterectomy [6
]--and inpatient hospitalization rates for general medical illnesses--such as back problems, gastroenteritis, and heart failure [7
]--have been shown to be related more strongly to the supply of a service than to the need for that service [8
]. While small area analysis has not helped health systems determine the optimal supply of health services, it is clear that small areas with the highest utilization rates experience the worst health outcomes even in the face of similar disease burdens [9
]. This has spurred concerns that an oversupply of health care may worsen health outcomes for a population [10
]. Chief among these concerns is that once the true need for a health service has been served, market forces dictate that excess supply must be consumed by those who do not actually need the services and are therefore exposed to the risk without the potential for benefit [2
]. While some conditions always merit treatment, others--so-called high-variation conditions--tend to be treated more intensively in the presence of excess resources [7
]. The development of the Dartmouth Atlas of Healthcare has facilitated the application of small area analysis to national datasets and allowed the identification of this phenomenon on a local level,[11
] as in reports of the over-provision of cardiac surgery in Reading, California [12
] and of unusually high rates of coronary stent procedures in Elyria, Ohio [14
]. Recent literature has been more critical of the concept of supply inducted demand in medicine. While there is general agreement that utilization and supply are correlated, there is less agreement regarding the meaning and drivers of this association [16
While small area analysis has been extensively applied to hospital-based medical and surgical services, there has been little application to hospital-based psychiatric services. A 1995 analysis of psychiatric inpatient admission patterns in Iowa found higher rates in small areas with more primary care physicians, psychiatrists, and inpatient psychiatric units [17
]. The authors concluded both that the differences were unlikely to be related to differences in need and that demand for inpatient psychiatry services was, in fact, sensitive to supply. However, there were several limitations to this analysis. First, the authors used standard hospital service areas (HSAs), which are based upon where most Medicare recipients who live in contiguous zip codes obtain general inpatient hospital services. As there are many more general hospitals than there are psychiatric hospitals and general hospitals with psychiatric units, most of the HSAs did not contain a psychiatric unit. This method did not allow researchers to account for use of inpatient psychiatric services in neighboring HSAs. An earlier analysis grouped these HSAs by county into politically-defined community mental health center (CMHC) catchment areas and found that access to CMHC resources induced demand for inpatient psychiatric admissions [18
]. However, the CMHC catchment areas were not necessarily the same or even intended to be the same as catchment areas for inpatient psychiatric units. Perhaps the most comprehensive study of geographic variation in inpatient mental health care examined county variation in New York [19
]. This study found that population variables such as poverty and population density were highly correlated with mental health service utilization. Furthermore, even when controlling for these factors, proximity to inpatient care resulted in increased utilization.
Often small area analysis has examined specialty care by aggregating HSAs into larger hospital referral regions (HRRs). HRRs are based upon where most Medicare recipients living in contiguous zip codes obtain heart surgery and neurosurgery [11
]. While useful for understanding geographic health service use patterns in expensive, highly technical procedures, these HRRs may not be as useful for understanding utilization of psychiatric inpatient services. As large inpatient psychiatric facilities are sometimes located in areas that do not have a medical referral hospital and as there is no analogous hierarchy of complexity in psychiatric units (for example, university-based psychiatric units do not necessarily offer more complex or technical procedures than community-based psychiatric units as is the case in medical hospitals), it is unlikely that geographic patterns of referral would be the same. A final concern with using standard HSAs is that patients admitted to inpatient psychiatric units tend to be younger than patients admitted to inpatient general medical units. As a result, fewer patients are eligible for Medicare; Medicare billing data may not be the most appropriate information to use to define HSAs in this population. Another major limitation of the Iowa and New York State studies is that they looked at only one state. As HSAs often cross state lines, it may be more meaningful to look at a region rather than a single state.
Another study examined geographic variation in inpatient psychiatric admission in New York City [20
]. The authors relied on zip codes as their unit of analysis and did not construct hospital service areas. They found that patients residing in a zip code where an inpatient psychiatric unit was located were more likely to be admitted. However, this analysis is subject to the same fallacy as the Iowa and New York State data, given that many zip codes did not have an inpatient psychiatric unit and that there is no reason to believe patients obtain their medical care within their zip code.
Therefore, if the intent is to study geographic variation in inpatient psychiatric admissions, it would be most helpful if: 1) each HSA--in this case Psychiatric HSAs or PHSAs--had at least one inpatient psychiatric hospital located within its boundaries, 2) we knew the capacity (number of psychiatric beds) of these inpatient psychiatric units rather than simply whether they existed or not, 3) we used the most population-relevant data (including using all adult age groups) to define our PHSAs, 4) we recognized that in some areas, especially along interstate borders, people are likely to live in one state and obtain healthcare in another, and 5) we parted from the assumption that these PHSAs will have some hierarchical regional organization as seen in general medical HSAs.
One of our goals in this study was to define the adaptations that would make small area analysis more relevant to the study of psychiatric care. We chose inpatient psychiatric treatment as a starting point both because of the large cost strain it places on the mental health treatment system and because of the previously-documented possibility of supplier-induced demand in this sector. Inpatient psychiatric treatment is an integral part of mental health treatment in the United States. In 2000, an estimated 215,221 inpatient psychiatric beds in 3,202 hospitals accommodated 2,153,874 psychiatric hospitalizations at a cost of almost 33 billion dollars [21
]. The large capacity of and costs associated with inpatient psychiatric care in the US reflects its central role in the provision of care for mental health patients. Overall, 74% of all mental health care dollars are spent on inpatient care: Medicare spends 83% of its mental health care budget on inpatient treatment,[22
] 65% of State mental health spending is for inpatient care, and Blue Cross/Blue Shield recently reported that 66% of their mental health care spending was for inpatient care [23
]. Despite the volume and cost of inpatient psychiatric care, there is very little research regarding its effectiveness. Indeed, no randomized clinical trials have demonstrated effectiveness for inpatient care in a general psychiatric population [24
We had three hypotheses in conducting this study, related to the actual subject at hand (inpatient psychiatric admissions) and the method to be used (applying small area analysis to mental health services). Our primary hypotheses were that there would be substantial variation in psychiatric admission across geographic settings and that bed availability would be positively correlated with admission rates, reflecting a supplier-induced demand phenomenon. Our secondary hypothesis was that the construction of psychiatric HSAs would yield more meaningful results than the use of existing general medical HSAs. This article reports the first small area analytic study of mental health services utilization using discipline-specific techniques. Our approach does not discount the considerable previous literature which has demonstrated other important drivers of inpatient mental health care including population factors (poverty and prevalence of mental illness), ambulatory treatment resources (availability and quality), and geographic proximity of inpatient service. Our focus on the quantity of inpatient mental health beds reflects the relative paucity of research regarding this variable compared to the important variables above.