ABSTRACT
BACKGROUND: Aggregated claims data on medication are often used as a proxy for the prevalence of diseases, especially chronic diseases. However, linkage between medication and diagnosis tend to be theory based and not very precise. Modelling disease probability at an individual level using individual level data may yield more accurate results. METHODS: Individual probabilities of having a certain chronic disease were estimated using the Random Forest (RF) algorithm. A training set was created from a general practitioners database of 276 723 cases that included diagnosis and claims data on medication. Model performance for 29 chronic diseases was evaluated using Receiver-Operator Curves, by measuring the Area Under the Curve (AUC). RESULTS: The diseases for which model performance was best were Parkinson's disease (AUC = .89, 95% CI = .77-1.00), diabetes (AUC = .87, 95% CI = .85-.90), osteoporosis (AUC = .87, 95% CI = .81-.92) and heart failure (AUC = .81, 95% CI = .74-.88). Five other diseases had an AUC >.75: asthma, chronic enteritis, COPD, epilepsy and HIV/AIDS. For 16 of 17 diseases tested, the medication categories used in theory-based algorithms were also identified by our method, however the RF models included a broader range of medications as important predictors. CONCLUSION: Data on medication use can be a useful predictor when estimating the prevalence of several chronic diseases. To improve the estimates, for a broader range of chronic diseases, research should use better training data, include more details concerning dosages and duration of prescriptions, and add related predictors like hospitalizations.
Subject(s)
Algorithms , Chronic Disease/drug therapy , Chronic Disease/epidemiology , Drug Utilization/statistics & numerical data , Drug Utilization/trends , Hospitalization/statistics & numerical data , Probability , Adult , Aged , Aged, 80 and over , Female , Forecasting , Humans , Male , Middle Aged , Netherlands/epidemiology , Population Surveillance/methods , PrevalenceABSTRACT
BACKGROUND: Since in an ageing society more long-term care (LTC) facilities are needed, it is important to understand the main determinants of first-time utilization of (LTC) services. METHODS: The Andersen service model, which distinguishes predisposing, enabling and need factors, was used to develop a model for first-time utilization of LTC services among the general population of the Netherlands. We used data on 214,821 persons registered in a database of general practitioners (NIVEL Primary Care Database). For each person the medical history was known, as well as characteristics such as ethnicity, income, home-ownership, and marital status. Utilization data from the national register on long-term care was linked at a personal level. Generalized Linear Models were used to determine the relative importance of factors of incident LTC-service utilization. RESULTS: Top 5 determinants of LTC are need, measured as the presence of chronic diseases, age, household size, household income and homeownership. When controlling for all other determinants, the presence of an additional chronic disease increases the probability of utilizing any LTC service by 45% among the 20+ population (OR = 1.45, 95% CI: 1.41-1.49), and 31% among the 65+ population (OR = 1.31, 95% CI: 1.27-1.36). With respect to the 20+ population, living in social rent (OR = 2.45, 95% CI = 2.25-2.67, ref. = home-owner) had a large impact on utilizing any LTC service. In a lesser degree this was the case for living alone (OR = 1.63, 95% CI = 1.52-1.75, ref. = not living alone). A higher household income was linked with a lower utilization of any LTC service. CONCLUSIONS: All three factors of the Anderson model, predisposing, enabling, and need determinants influence the likelihood of future LTC service utilization. This implies that none of these factors can be left out of the analysis of what determines this use. New in our analysis is the focus on incident utilization. This provides a better estimate of the effects of predictors than a prevalence based analysis, as there is less confounding by changes in determinants occurring after LTC initiation. Especially the need of care is a strong factor. A policy implication of this relative importance of health status is therefore that LTC reforms should take health aspects into account.
Subject(s)
Nursing Homes/statistics & numerical data , Activities of Daily Living , Adult , Aged , Aged, 80 and over , Chronic Disease , Decision Making , Female , Health Status , Humans , Long-Term Care/statistics & numerical data , Male , Middle Aged , Netherlands , Primary Health Care , Young AdultABSTRACT
Health care utilization is expected to rise in the coming decades. Not only will the aggregate need for health care grow by changing demographics, so too will per capita utilization. It has been suggested that trends in health care utilization may be age-specific. In this paper, age-specific trends in health care utilization are presented for different health care sectors in the Netherlands, for the period 1981-2009. For the hospital sector we also explore the link between these trends and the state of medical technology. Using aggregated data from a Dutch health survey and a nationwide hospital register, regression analysis was used to examine age-specific trends in the probability of utilizing health care. To determine the influence of medical technology, the growth in age-specific probabilities of hospital care was regressed on the number of medical patents while adjusting for confounders related to demographics, health status, supply and institutional factors. The findings suggest that for most health care sectors, the trend in the probability of health care utilization is highest for ages 65 and up. Larger advances in medical technology are found to be significantly associated with a higher growth of hospitalization probability, particularly for the higher ages. Age-specific trends will raise questions on the sustainability of intergenerational solidarity in health care, as solidarity will not only be strained by the ageing population, but also might find itself under additional pressure as the gap in health care utilization between elderly and non-elderly grows over time. For hospital care utilization, this process might well be accelerated by advances in medical technology.
Subject(s)
Biomedical Technology/statistics & numerical data , Health Services/statistics & numerical data , Adult , Age Distribution , Age Factors , Aged , Aging , Female , Hospitals/statistics & numerical data , Humans , Male , Middle Aged , Netherlands , Patents as Topic/statistics & numerical data , Sex DistributionABSTRACT
BACKGROUND: Considering the scarcity of health care resources and the high costs associated with cardiovascular diseases, we investigated the spending on cardiovascular primary preventive activities and the prescribing behaviour of primary preventive cardiovascular medication (PPCM) in Dutch family practices (FPs). METHODS: A mixed methods design was used, which consisted of a questionnaire (n = 80 FPs), video recordings of hypertension- or cholesterol-related general practitioner visits (n = 56), and the database of Netherlands Information Network of General Practice (n = 45 FPs; n = 157,137 patients). The questionnaire and video recordings were used to determine the average frequency and time spent on cardiovascular primary preventive activities per FP respectively. Taking into account the annual income and full time equivalents of general practitioners, health care assistants, and practice nurses as well as the practice costs, the total spending on cardiovascular primary preventive activities in Dutch FPs was calculated. The database of Netherlands Information Network of General Practice was used to determine the prescribing behaviour in Dutch FPs by conducting multilevel regression models and adjusting for patient and practice characteristics. RESULTS: Total expenditure on cardiovascular primary preventive activities in FPs in 2009 was 38.8 million (2.35 per capita), of which 47% was spent on blood pressure measurements, 26% on cardiovascular risk profiling, and 11% on lifestyle counselling. Fifteen percent (11 per capita) of all cardiovascular medication prescribed in FPs was a PPCM. FPs differed greatly on prescription of PPCM (odds ratio of 3.1). CONCLUSIONS: Total costs of cardiovascular primary preventive activities in FPs such as blood pressure measurements and lifestyle counselling are relatively low compared to the costs of PPCM. There is considerable heterogeneity in prescribing behaviour of PPCM between FPs. Further research is needed to determine whether such large differences in prescription rates are justified. Striving for an optimal use of cardiovascular primary preventive activities might lead to similar health outcomes, but may achieve important cost savings.
Subject(s)
Cardiovascular Diseases/economics , Cardiovascular Diseases/prevention & control , Chemoprevention/economics , Family Practice , Primary Prevention/economics , Female , Humans , MaleABSTRACT
PURPOSE: Off-hours admissions have been associated with an increased risk of poor outcomes but results have been inconsistent, possibly due to different measures of off-hours care used. We examined, using a single condition and increasingly refined definitions of time of admission, the effect of off-hours admissions on 7-day stroke case-fatality. METHODS: We studied a retrospective cohort of 82,219 ischaemic stroke admissions to 115 Dutch hospitals between 2000 and 2004. Data were from the Dutch Medical Register and analysed using multivariable multilevel logistic regression. We adjusted for variables such as age, gender, Charlson-Deyo comorbidity score, urgency of admission, hospital teaching status and speciality of attending physician. RESULTS: After adjustment, we observed higher 7-day death risk for weekend admissions when compared to weekday admissions (OR 1.27; 95% CI 1.20 to 1.34). Sunday displayed the highest risk of death (OR 1.31; 95% CI 1.20 to 1.44). With the Monday day-shift as a reference, the death odds were increased during the Sunday and Saturday day-shifts, the evening-shifts on Sunday and Monday, and during all night-shifts. The night-shift ORs ranged from 1.94 (95% CI 1.56 to 2.41) to 2.14 (95% CI 1.74 to 2.63). When compared to admission at 8:00 we observed increased death odds from midnight until 7:00 and decreased death odds from 14:00 until 18:00. CONCLUSIONS: Weekends represent a period of increased death risk for ischaemic stroke patients in the Netherlands. However, this increased risk appears to represent an exacerbation of an underlying night-time risk present during the weekdays.
Subject(s)
Brain Ischemia/mortality , Stroke/mortality , Age Factors , Aged , Brain Ischemia/complications , Cohort Studies , Ethnicity , Female , Hospitals , Humans , Logistic Models , Male , Netherlands/epidemiology , Patient Admission , Personnel Staffing and Scheduling , Retrospective Studies , Social Class , Stroke/etiology , Time Factors , Treatment OutcomeABSTRACT
A shortcoming of many economic evaluations is that they do not include all medical costs in life-years gained (also termed indirect medical costs). One of the reasons for this is the practical difficulties in the estimation of these costs. While some methods have been proposed to estimate indirect medical costs in a standardized manner, these methods fail to take into account that not all costs in life-years gained can be estimated in such a way. Costs in life-years gained caused by diseases related to the intervention are difficult to estimate in a standardized manner and should always be explicitly modelled. However, costs of all other (unrelated) diseases in life-years gained can be estimated in such a way. We propose a conceptual model of how to estimate costs of unrelated diseases in life-years gained in a standardized manner. Furthermore, we describe how we estimated the parameters of this conceptual model using various data sources and studies conducted in the Netherlands. Results of the estimates are embedded in a software package called 'Practical Application to Include future Disease costs' (PAID 1.0). PAID 1.0 is available as a Microsoft® Excel tool (available as Supplemental Digital Content via a link in this article) and enables researchers to 'switch off' those disease categories that were already included in their own analysis and to estimate future healthcare costs of all other diseases for incorporation in their economic evaluations. We assumed that total healthcare expenditure can be explained by age, sex and time to death, while the relationship between costs and these three variables differs per disease. To estimate values for age- and sex-specific per capita health expenditure per disease and healthcare provider stratified by time to death we used Dutch cost-of-illness (COI) data for the year 2005 as a backbone. The COI data consisted of age- and sex-specific per capita health expenditure uniquely attributed to 107 disease categories and eight healthcare provider categories. Since the Dutch COI figures do not distinguish between costs of those who die at a certain age (decedents) and those who survive that age (survivors), we decomposed average per capita expenditure into parts that are attributable to decedents and survivors, respectively, using other data sources.
Subject(s)
Health Care Costs/standards , Health Expenditures/standards , HumansABSTRACT
BACKGROUND: A few studies have found an inverse association between hospital patient volume and case-fatality among stroke patients. However, the different stroke categorizations used in these studies might have influenced the findings. Furthermore, the relevance of the association observed remains questionable given that the relatively small magnitude may not support volume-based referral policies. We re-examined this association in a large nationwide study, paying attention to the influence of volume categorizations. METHODS: Applying multilevel logistic regression, we re-examined the relationship between hospital stroke volume and 7-day case-fatality using admissions data obtained from Statistics Netherlands on 73,077 stroke patients for the years 2000 to 2004. Different cut-offs were used to categorize hospitals in volume groups. We also examined the implications of a volume based referral strategy. RESULTS: Stroke patients in high-volume hospitals had decreased risk of dying within 7 days of admission even when different hospital categorizations are applied. For instance, the odds ratio was 0.45(95% CI 0.20-0.99) in high-volume(>200 case-volume) versus low-volume(<50 case-volume) hospitals, but 0.89(95% CI 0.79-1.00) in high-volume(>250 case-volume) versus low-volume (< or =250 case-volume) hospitals. Ignoring travel time and workload implications an optimistic volume-based referral policy would save 183 patients when all patients are referred to the >200 case-volume hospital. A nontransfer policy aimed at reducing mortality by 10% in all those hospitals would save 1260 patients. CONCLUSION: Stroke patients in low-volume versus high-volume hospitals have higher odds of dying. This finding may not lend itself to a substantial volume-based referral strategy.
Subject(s)
Outcome and Process Assessment, Health Care , Patient Admission/statistics & numerical data , Patient Transfer , Stroke/mortality , Aged , Aged, 80 and over , Female , Hospital Mortality , Humans , Logistic Models , Male , Middle Aged , Multivariate Analysis , Netherlands/epidemiology , Retrospective Studies , Stroke/classificationABSTRACT
BACKGROUND: Patterns in time, place and cause of death can have an important impact on calculated hospital mortality rates. Objective is to quantify these patterns following myocardial infarction and stroke admissions in Dutch hospitals during the period 1996-2003, and to compare trends in the commonly used 30-day in-hospital mortality rates with other types of mortality rates which use more extensive follow-up in time and place of death. METHODS: Discharge data for all Dutch admissions for index conditions (1996-2003) were linked to the death certification registry. Then, mortality rates within the first 30, 90 and 365 days following admissions were analyzed for deaths occurring within and outside hospitals. RESULTS: Most deaths within a year after admission occurred within 30 days (60-70%). No significant trends in this distribution of deaths over time were observed. Significant trends in the distribution over place of death were observed for both conditions. For myocardial infarction, the proportion of deaths after transfer to another hospital has doubled from 1996-2003. For stroke a significant rise of the proportion of deaths outside hospital was found. For MI the proportion of deaths attributed to a circulatory disease has significantly fallen overtime. Seven types of hospital mortality indicators, different in scope and observation period, all show a drop of hospital mortality for both MI and stroke over the period 1996-2003. For stroke the observed absolute reduction in death rate increases for the first year after admission, for MI the observed drop in 365-day overall mortality almost equals the observed drop in 30-day in hospital mortality over 1996-2003. CONCLUSION: Changes in the timing, place and causes of death following admissions for myocardial infarction and stroke have important implications for the definitions of in-hospital and post-admission mortality rates as measures of hospital performance. Although necessary for understanding mortality patterns over time, including within mortality rates deaths which occur outside hospitals and after longer periods following index admissions remain debatable and may not reflect actual hospital performance but probably mirrors transfer, efficiency, and other health care policies.