Impact of Inappropriate Drug Use on Hospitalizations, Mortality, and Costs in Older Persons and Persons with Dementia: Findings from the SNAC Study

Data for this study were obtained from the ongoing, longitudinal cohort study Swedish National Study on Aging and Care (SNAC), which is described in detail elsewhere [12]. In short, SNAC is a study conducted at four sites in Sweden (Nordanstig, Kungsholmen, Karlskrona, and four municipalities in Skåne), reflecting both rural and urban parts of Sweden. People aged 60 years and over in certain age groups (60, 66, 72, 78, 81, 84, 87, 90, 93, 96, 99 years) were recruited for an extensive examination, which included a physical examination and medical history by a physician. Interviews covering a wide range of areas such as socioeconomic factors, physical environment, and care service utilization were also conducted by a nurse, and tests of psychological status and cognitive performance were made by a psychologist. The data used in this study were collected at baseline during the years 2001–2004 in Nordanstig, a rural area in the middle of Sweden, and Kungsholmen, an urban area in Stockholm (n = 4129).

Age was categorized into 60–69, 70–79, 80–89, and ≥90 years in the descriptive analysis and used as a continuous variable in the regression models. Residential setting was dichotomized into community dwelling (i.e., living in one’s own home) versus institution (sheltered accommodation, old people’s home, group dwelling, or nursing home). Educational level was classified into elementary versus additional schooling.

Use of drugs was recorded by the physician and the participants were asked to bring current lists of medications, drug containers, and prescriptions. If the participant was not able to answer, an informant close to the person (relative or carer) was asked to provide the information. Drug use was defined as use of a drug regularly at the time of the interview or as needed at any time during the preceding month. Regular use was defined as the use of a drug on a scheduled basis (e.g., day, week, or fortnight). Data on both prescribed drugs and over-the-counter drugs were collected [12]. The dose and frequency of the drugs used were also collected. Polypharmacy was defined as use of five or more drugs regularly. Drugs were classified according to the anatomical therapeutic and chemical (ATC) system [13].

IDU was defined according to the indicators developed by the Swedish National Board of Health and Welfare [5, 14, 15]: concurrent use of three or more psychotropic drugs (from any of the groups antipsychotics, anxiolytics, hypnotics-sedatives, and antidepressants), use of drugs with anticholinergic properties (urinary and gastrointestinal antispasmodics, anticholinergic antiemetics, class Ia antiarrhythmics, anticholinergic antiparkinsonian drugs, low-potency antipsychotics, tricyclic antidepressants, and first-generation antihistamines), use of long-acting benzodiazepines (diazepam, nitrazepam, or flunitrazepam), and serious drug–drug interactions according to the system by Sjöqvist [16]. IDU was defined as exposure to at least one of these four indicators [5].

Dementia was diagnosed by physicians according to the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R) [17].

We used the Charlson Comorbidity Index, which is often used to adjust for confounding by co-morbidities [18]. The version applied in this study was adapted to the availability of data [8]. We used seven diagnoses with a weight of one (congestive heart failure, myocardial infarction, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disorder, and diabetes mellitus without complication) and two diagnoses with a weight of two (moderate or severe renal failure and any tumor), which gives a maximum total of 11. The comorbidity index was entered as a dichotomous variable in the analysis (none vs. one or more modified Charlson Comorbidity Index weights).

Diagnoses of diseases in the index were all based on medical history and examination of the physician except renal failure and dementia. Renal failure was calculated from Cockcroft-Gault formula [19] and defined as estimated creatinine clearance <25 mL/min. Dementia diagnosis is included in the Charlson Comorbidity Index, but dementia was analyzed as a separate variable in the regression analysis and was, thus, removed from the index.

Physical functioning was assessed by the Katz index of activities of daily living (ADL) [20]. The Katz ADL is a scale that measures functional dependency in six basic activities: transferring, dressing, bathing, going to the toilet, feeding, and continence. Functional independence was defined as no need of assistance and dependent was defined as being dependent in one or more activities.

Data on mortality and hospitalizations were collected from registers at the Swedish National Board of Health and Welfare. Both death certificates and data on hospitalization were collected from individual entrance into the study until 1 year after the last date of inclusion in the study.

Cost data of hospitalization were derived from the Nord-Diagnose Related Group (DRG) cost database [21]. This is a Swedish version of the original Diagnosis Related Group (DRG) database [22], which was produced to rationalize cost-finding for budgeting and visualize costing for practitioners. Each DRG code was derived from the corresponding International Classification of Diseases (ICD) code [23] with a specific weight. This weight was then multiplied by the DRG cost of weight 1 in order to calculate the total cost of the hospitalization.

Any health economic analysis needs to have a specified viewpoint [24]. This paper focuses on costs of hospitalizations, which is a part of medical sector costs. Costs of the community care sector and costs of informal care are therefore not included. Furthermore, the only additional outcome was mortality. No other effects, such as quality of life or functional capacity, were included. The basic health economic design is descriptive and there is no predefined hypothesis regarding cost effectiveness. Thus, the health economic approach in this paper can be regarded as a cost description that can generate hypotheses for interventions regarding IDU.

Descriptive demographic statistics were made with cross-tabulations. Logistic regression analysis was used to investigate the association between IDU and hospitalizations within 1 year of assessment of IDU, after adjustment for covariates. Similarly, Cox regression models were used for analysis of IDU and mortality within 1 year of assessment of IDU, after adjustment for covariates. The outcomes hospitalizations and mortality were analyzed in the whole study population and in the subpopulation of persons with dementia. There were no significant differences in IDU between the two sites, Nordanstig and Kungsholmen, and inclusion of site did not affect the results of the regression models; thus, this variable was not included in the analyses. The results are shown as odds ratios (ORs) and hazard ratios (HRs) with 95 % confidence intervals (95 % CIs). We used one-way analysis of variance (ANOVA) to explore differences in mean cost of hospitalizations with and without IDU, after adjustment for age. A p value of <0.05 was considered statistically significant. All analysis was made with IBM SPSS^® Statistics version 22 [25].

The study was approved by the ethical review board in Stockholm (dnr 01-114, dnr 2009/595-32) and in Uppsala (dnr 01-123). Informed consent was obtained from all participants included in the study.