Measuring the costs of primary care education in the ambulatory setting.

In 1995, the authors obtained cost, operations, and educational activity data from 98 ambulatory care sites across the United States in which primary care teaching was occurring and compared those data with the corresponding data from 84 ambulatory care sites where no teaching was going on. The teaching sites in the sample were found to have 24-36% higher operating costs than the non-teaching sites. This overall difference in costs is approximately the same difference in costs earlier estimated for university teaching hospitals compared with non-teaching hospitals. These costs are shared by all involved in the ambulatory education process: sponsors, sites, and faculty. In a related finding, the authors discovered that 30-50% of all ambulatory care sites thought not to be involved in education are in fact teaching at a high level of involvement. Further research into not only the costs but the value of education in the clinical setting is encouraged. The authors also hope that the publication of this report will encourage accrediting bodies and professional organizations to improve the information available about ambulatory care training in general.

Knowledge of mental health benefits and preferences for type of mental health providers among the general public.

OBJECTIVE: The study explored knowledge of mental health benefits and preferences for providers among the general public. METHODS: Analysis was based on a telephone survey of 1,358 adults randomly sampled throughout Michigan in 1997-1998. RESULTS: A large proportion of the respondents were uninformed about their mental health benefits. One-quarter of the sample were unsure if their health plan even included mental health services. Forty-three percent of the sample believed that mental health benefits were equal to benefits provided for general medical services. In answer to a survey question that summarized payment restrictions for psychiatric services and counseling under Medicare, nearly a quarter of older respondents indicated that they would not seek care even when needed. In the overall sample, the majority of respondents said they would initially seek care from their primary care physician for a mental health problem, although responses varied by age. Persons over age 65 were significantly more likely to seek assistance from their primary care doctor than were younger persons. CONCLUSIONS: The general public lacks information about important mental health benefits, and this lack of information may represent a barrier in their seeking care when needed. Given the overriding preference for primary care providers to treat mental health problems, particularly among older adults, mental health issues should be given more attention at all levels of primary care education.

Estimating the cost of primary care training in ambulatory settings.

The Balanced Budget Act of 1997 legislated the idea of reimbursing ambulatory sites for training medical professionals. However, very little is known about the costs of training in such settings. This paper assesses the cost of primary care training in ambulatory settings. Selection models were used to separate the cost of teaching from the cost of infrastructural differences between teaching and non-teaching sites. A probit equation modelled the likelihood of an ambulatory site having a teaching programme and a cost function related total medical practice costs to clinical output, the presence of a health professions educational programme, the price of resources used, characteristics of the medical practice and location. Data on 184 community health centres (CHCs), group practices, health maintenance organizations (HMOs) and outpatient clinics were used. Teaching sites were found to have 36% higher operating costs than their non-teaching counterparts: 38% of these higher costs were due to infrastructural differences and 62% were the 'pure' costs of teaching, i.e. the costs of teaching the net of infrastructural effects.

When (not if) evaluation flexibility is desirable. Examples from the CPHPE initiative.

The evaluation literature often debates whether evaluators should be flexible in evaluation design and activities in order to collaborate with program directors and be responsive to programming needs. Two conditions are specified under which evaluation flexibility is not only desirable but essential. Two examples from the cluster evaluation of the W. K. Kellogg Foundation's Community Partnerships for Health Professions Education initiative are provided to illustrate why flexibility under these conditions proved to be essential. One of the examples, related to the "community" involvement in the initiative, illustrates the need for flexibility as programs experience goals clarification. The other example, related to the coincidental national health care reform efforts, illustrates the need for flexibility both to capture programs' efforts to protect their integrity and to ensure against spurious conclusions as a result of external turbulence in policy environments. How the cluster evaluation team addressed these issues is also described.

Methodological issues in linking costs and health outcomes in research on differing care delivery systems.

OBJECTIVES: This article discusses, from an economist's point of view, issues in designing and conducting research including cost and outcomes variables among differing care delivery systems. METHODS: Issues were identified, and selected research purporting to link cost and outcomes with variations in care delivery systems was reviewed. RESULTS: Current literature on nursing care delivery systems and costs in hospitals, ambulatory care, and nursing homes is focused mainly on group-specific costs linked to patient-specific outcomes. It suffers further from focusing primarily on single-discipline components of care, omitting the contributions of other providers beyond nursing personnel. CONCLUSIONS: Multidisciplinary teams, including economists who can speak clinical language, are recommended.

HIV infection treatment costs under Medicaid in Michigan.

The Michigan Medicaid Program payment records generated in the period 1985-89 by 783 persons were analyzed for services related to human immunodeficiency virus (HIV) infection. Other data from death records and the Michigan AIDS Surveillance Registry were available for a subset of those persons. The average monthly payment in 1989 dollars for HIV-related services was $1,302.57. Services determined to be unrelated to HIV infection accounted for 12.5 percent of the total amount for health care received and another 2.5 percent was questionable. The average monthly expenditure for men was roughly twice that for women. The discrepancy did not exist among persons identified in the AIDS Surveillance Registry. Sex differences ceased to exist when Medicaid eligibility (disability versus Aid to Families with Dependent Children) was controlled for by analysis of variance. There were no significant differences between payments to those infected through male-to-male sexual contact and those infected through intravenous drug use. Payments for HIV treatments rose with age to about 40 years, and declined slightly among older adults. The sharpest rise was for those ages 19-25 years and 26-35 years. Large sex differences existed among those who received zidovudine (AZT), 61.4 percent of the men and 19.1 percent of the women. Controlling for Medicaid eligibility moderated those differences, but they remained statistically significant. Differences in zidovudine usage were not found between men and women in the subset identified in the AIDS Surveillance Registry nor among persons infected through male-to-male sexual contact and intravenous drug use.

Changes in rates of unscheduled hospital readmissions and changes in efficiency following the introduction of the Medicare prospective payment system. An analysis using risk-adjusted data.

The purpose of this study was to analyze changes in rates of unscheduled readmissions and changes in technical efficiency following the introduction of the Medicare Prospective Payment System (PPS). We developed the Risk-Adjusted Readmissions Index (RARI), which allowed us to make comparisons in rates of unanticipated readmissions across hospitals and over time. Data envelopment analysis (DEA), a linear programming technique, was used to measure changes in technical efficiency by comparing the inputs used and the outputs produced across a cohort of hospitals, while adjusting for changes over time in case mix and case complexity. Rates of unscheduled readmissions and efficiency scores were computed for a sample of 245 hospitals for each year. Although both readmission rates and efficiency scores increased for most hospitals, there was no evidence that those hospitals that experienced the greatest increases in efficiency had the largest increases in their rates of unscheduled readmissions.

Measuring hospital performance. The development and validation of risk-adjusted indexes of mortality, readmissions, and complications.

In this study we used information from discharge abstracts to develop three different risk-adjusted measures of hospital performance: a Risk-Adjusted Mortality Index, a Risk-Adjusted Readmissions Index, and a Risk-Adjusted Complications Index. The adjustments have face validity, and appear to account for much of the variation across hospitals in the rates of these adverse events. The indexes are stable over time, and are not biased with respect to hospital size, ownership, or teaching status. All three indexes appear to have construct validity when tested against the changes in hospital care that occurred when PPS was introduced.

Cholesterol measurement and treatment in community practices.

A study was designed to examine the cholesterol measurement and treatment activities of primary care physicians in community practices. Three family practices of comparable size (one faculty practice and two community small-group practices) participated in the study. A random sample of 450 adult patients (150 from each site) was drawn from patient logs using a time series sampling method. Charts were reviewed for serum lipid evaluations, documentation of coronary heart disease risk factors, lipid-lowering activities, and other coronary heart disease risk-factor interventions. Sixty-seven percent of the sample had cholesterol measures recorded. No differences were found in the rates of measurement for men and women. Multiple, detailed serum lipid evaluations were common, and recognition of high cholesterol as a problem even before 1980 was apparent. Almost one half (47%) of individuals with cholesterol greater than 5.2 mmol/L (200 mg/dL) had a charted intervention, 64% if cholesterol greater than 6.2 mmol/L (240 mg/dL). Diet was the most common intervention (73%), and medication was used in only eight cases. Nonpharmaceutical interventions appeared to be undercharted. An analysis of interpractice variations revealed strikingly consistent results, although some interesting differences were noted. These rates are at least double previously reported rates and suggest that primary care physicians play a major role in this national priority.

How to improve allocation of support service costs.

Better accounting for support service costs at outpatient facilities involves distinguishing between fixed and variable expenses, then creating separate budgets for them. To simplify this step, financial managers can create a surrogate, such as the number of patient visits, to represent service activity. Once this is completed, separated costs are allocated by using criteria that recognize short-term and long-term service use.

Analysis of Michigan Medicaid costs to treat HIV infection.

To obtain better understanding of the nature and cost of health care related to human immunodeficiency virus (HIV) infection, medical payment records were analyzed for 204 men, women, and children older than 60 months who had indications of HIV infection. The records were those of Michigan Medicaid, the General Assistance Medical Program, and the Resident County Hospitalization Program, with service dates on or after January 1, 1984, and which were processed by November 30, 1987. Patient payment records were coded according to whether the patient's condition was considered to be pre-HIV, HIV unrelated, possibly HIV related, or HIV related. Average monthly payments were found to be $150 for pre-HIV patient payment records, $114 for those HIV unrelated, $57 for those possibly related, and $1,213 for those related to HIV infection. HIV-related monthly payments rose from about $1,500 per month in the period 3 months prior to the patient's death to more than $8,000 in the last month of life. Men were found to have twice as many claims as women, and men's claims cost about three times as much. A higher percentage of women than men (91 percent versus 37 percent) received pre-HIV paid services, indicating a higher percentage of women were at least initially receiving Medicaid for reasons other than an HIV-related disability. Diagnostic categories that accounted for the bulk of the HIV-related health care utilization included infectious and parasitic diseases, acquired immunodeficiency syndrome, diseases of the respiratory system, and non-HIV-specific immunity disorders. Inpatient hospitalization accounted for more than 75 percent of the payments, followed by physician costs (11 percent), pharmacy costs (5 percent), and outpatient costs (3 percent). A total of 45, or about 22 percent of the recipients, received zidovudine (AZT) prescriptions at an average monthly cost of $404.

The effects of nursing and physician services: some preliminary results.

This paper examines the relationship between the future staff cost of nursing home patient care and the type and level of nursing and medical services these patients receive. Nursing times are grouped into eight categories based on a principal components analysis. We then investigate the effect of the level and type of nursing care and of physicians' services given in an earlier period on the staff cost of nursing home care in a later period. Psychosocial nursing care (emotional support, education) and number of physicians' visits given in the earlier period were found to be negatively associated with costs in a later period. The most powerful predictor of future costs were actual costs in the earlier period. These results suggest that more investment in psychosocial nursing effort and physicians' services may yield modest cost savings in future nursing home patient care.

Patient classification and resource allocation in Veterans Administration nursing homes.

Patient classification for Veterans Administration and non-Veterans Administration long-term care patients is discussed. Results are reported from 290 patients in two VA nursing homes by using both resource utilization groups and an independently generated classification system. Patient classes were generated by using the Automatic Interaction Detection Program, which was the methodology used to create the diagnosis related groups. The use of diagnoses in long-term care patient classification is also reviewed. The effect of disaggregated nursing times by type of provider and by type of nursing activity on patient classes is examined.

Activities of daily living as quantitative indicators of nursing effort.

Functional assessments of elderly or disabled people requiring long-term care have been used by clinicians for many years, and functional assessment instruments are now being used as indicators of required nursing care and its cost. The authors examine the ability of functional assessment items and instruments to measure accurately the variation in nursing care used by nursing home patients, with analysis of 290 patients. Nursing times, measured for each patient by nurse category (registered and all other) and type of care (skilled and personal) measure resource consumption. Activities of daily living (ADLs): eating, bathing, dressing, toileting, transferring, and continence are used to measure functional abilities on a four-point scale: independent, supervised, assisted or helped, and dependent, as well as two derived scales: Katz's Index and Resource Utilization Groups. The four-point measurement scales for ADLs are found to be necessary as indicators of nursing time required by patients. As a consequence, the three-point scales used for ADLs in the Long-term Care Minimum Data Set are not adequate, at least in nursing homes, for resource allocation. The relationship of nursing times with individual ADLs is nonlinear, so linear statistical techniques such as principal components, canonical correlations, or linear regression are inappropriate to produce patient classification systems based on ADLs. Individual ADLs do not explain use of registered nursing care time as well as they do care time by other nursing staff. Therefore, resource allocation and staffing for registered nurses must be done separately from nursing personnel, using indicators other than ADLs.

Modeling the outcomes of nursing home care.

In this exploratory analysis using data on 290 patients, we use regression analysis to model patient outcomes in two Veterans Administration nursing homes. We find resource use, as measured with minutes of nursing time, to be associated with outcomes when case mix is controlled. Our results suggest that, under case-based reimbursement systems, nursing homes could increase their revenues by withholding unskilled and psychosocial care and discouraging physicians' visits. Implications for nursing home policy are discussed.

Capital expenditure planning: the value of information to hospitals.

This article examines some of the costs caused by the uncertainty created by competition among hospitals, especially those costs reflected in capital expenditure planning. A proposal is developed for a computer-based capital budgeting system that would allow hospitals to share enough information about their future capital expenditure plans to reduce the major costs of competition-induced uncertainty without publicly disclosing their future intentions. The benefits of such a system can be measured as the expected value of complete information. An illustrative example is provided.

Classifying nursing home patients; the impact of mid-week data collection on variance reduction.

This study analyzes the variance reduction capabilities of three existing nursing home patient classification systems across different shifts and sub-week time periods. The existing classification systems, all based on nursing time data collected in the middle of the working week, performed well with data collected at night and on weekends. Existing nursing home patient classification systems are robust for time periods other than those during which nursing times have usually been collected.

Long-term care patient assessment. Evaluation of a methods effect.

Evaluation of a long-term care patient classification instrument in use in the Veterans Administration revealed that the levels of functional ability used in such instruments can influence their abilities to explain use of nursing time and reliability between raters. Specifically, a three-level scale that combines patients who require supervision with those who require help does not explain use of LPN and nurse's aide personal care time as well as does a four-level scale or a three-level scale that combines patients who require supervision with those who are independent. These findings should be considered in designing and applying patient assessment instruments.

Functional assessment of nursing home patients. Reliability and relevance.

Functional assessment of patients for resource allocation or staffing requires a higher level of interrater reliability than functional assessment in normal clinical settings. Although many functional assessment instruments are available, interrater reliability of these items has frequently not been reported. An assessment instrument based on the Long Term Care Minimum Data Set format was used for 290 patients in six wards in two Veterans Administration nursing homes. Each patient was assessed independently by two nurse caregivers to obtain reliability information. Different reliability measures yielded differing evaluations of the reliability of the instrument. Absolute agreement rates combined with Kendall's tau-b were most useful in deciding on the reliability of items in the instrument.