Potentially inappropriate prescription of drugs in hospitalized older patients

Resumen

Potentially inappropriate prescriptions (PIP) are frequently found in health care settings and can generate negative outcomes such as adverse effects, readmissions and higher cost; however they are potentially avoidable (Dalleur et al. 2012; Nyborg et al., 2012; Laroche et al, 2007; Dedhiya et al., 2010; Bradley et al., 2012; Wenger et al., 2001). Older people are particularly susceptible because they often suffer from more than one chronic disease. Moreover, the time of discharge represents a high-risk situation, as new drugs are almost always prescribed which may be potentially inappropriate (Parsons et al., 2002; Hu et al., 2012; American College of Clinical Pharmacy et al., 2012). PIP in older patients at hospital discharge are frequent and should be a concern in Spanish hospitals (López-Sáez et al., 2012; Blasco-Patiño et al., 2008); however by intervening in the factors associated with them (patient-related or drug-related) the majority of PIP might be avoided. To reduce the frequency of this potential hazard, it is of great importance to identify the underlying causes and determine when it tends to occur most. This means that a reliable, reproducible and valid instrument of measurement is essential. Several criteria, mostly explicit, have been developed to detect potentially inappropriate prescriptions (PIP) and help physicians to optimize pharmacotherapy in elderly people (Spinewine et al., 2009). Among these, Beers (2012 update) and STOPP criteria are the most widely applied and studied (American Geriatrics Society 2012 Beers criteria update expert panel, 2012; Lam et al., 2012; Ryan et al., 2013). Despite the availability of these tools to improve prescribing, there is scarce data for a clear relationship between PIP and health outcomes which would ensure the validity of PIP measurement. It is assumed that PIP are associated with increased complications and adverse events, as well as overuse of health services. Despite this, only few studies (Kanaan et al., 2013; Reich et al., 2014; Tosato et al., 2014) have addressed this association and results are controversial or inconclusive. Our specific objectives were: 1. To measure the frequency of PIP in older people (¿ 65 years old) at hospital discharge identified by two different tools (Beers and STOPP criteria) 2. To analyze the association of PIP with different predictive factors (age, gender, Charlson Comorbidity Index, number of drugs prescribed at discharge, pharmacological group of each drug prescribed, the Hospital Specialty, length of hospital stay etc.) 3. To identify the drugs most commonly involved in PIP according to both criteria. 4. To identify the patients¿ characteristics that may influence the occurrence of PIP 5. To study the association between PIP in older patients at hospital discharge and health outcomes measured in the short to medium term, particularly mortality, number of hospital readmissions, primary care consultations, home visits and emergency treatment recorded from the date of hospital discharge to the last contact with the health system. Cross sectional study. The reference population consisted of older people (65 years and more) discharged from the University Specialty Hospital San Cecilio, Granada, during the period from July 1, 2011 to June 30, 2012. The hospital belongs to the Andalusian Health Service, serving the Granada Midwest Hospital area population of 346,682 inhabitants (about 50,962 older than 64 years). Inclusion criteria were: ¿ Presenting a discharge between July 2011 and June 2012, both inclusive, from the services of surgery, traumatology, internal medicine and other medical specialties (including cardiology, gastro-intestinal, respiratory, endocrinology and nephrology). ¿ 65 years old or more ¿ Charlson Comorbidity Index ¿2 (obtained through personal history and diagnoses reflected in the hospital discharge report) Exclusion criteria were: ¿ Discharge due to transfer to another hospital ¿ Discharge due to death ¿ Discharge from the services of dermatology, ophthalmology and otorhinolaryngology. The documentation service provided a list of patients 65 years old or more discharged during the study period for whom the reason of discharge was different from death or transfer to another hospital (N = 8,154). A 15 % random sample of this group (N = 1,004) was drawn using the STATA resampling option, in order to warrant 3 % accuracy after dropout patients who did not meet inclusion criteria; 361 (40.0 %) patients were excluded because the Charlson Comorbidity Index was lower than 2 and another 19 patients (1.9 %) because they were discharged from any one of the excluded services. The final sample therefore included a total of 624 patients that fulfilled the inclusion criteria. This number allowed us to estimate a proportion of 20 % with an accuracy of 3 %, and arrive at significant differences for a 6 percentage point difference with a minimum power of 80 %. Information Sources: ¿ The Hospital database, provided through the Documentation Service for the period of study, gave each patient¿s history number, age, gender, dates of admission and discharge, principal diagnosis and Diagnosis Related Group (DRG) code. ¿ Clinical history: Discharge report. If information available from these sources was insufficient, the pharmacist consulted the discharge report. The initial assessment including the review of patients¿ hospital discharge information was conducted between October 2012 and February 2013. We also collected information after discharge between December 2012 and March 2013 by consulting the database of the Clinical History Diraya* for these patients. Independent study variables were: age; gender; Charlson Comorbidity Index (from 2 to 37, calculated using the calculator type.xls of Hall) (Hall et al., 2004); number of drugs prescribed at discharge; pharmacological group of each drug prescribed; the Hospital Specialty (we grouped the cases into three categories: surgery, internal medicine and other medical services); length of hospital stay and DRG. Dependent variables were: number of PIP according to the two criteria used and number of patients with at least one PIP in their treatment plan, according to the two criteria used (Beers 2012 update and STOPP). Using Excel 2007 we recorded the prescription of each drug included under either of the two criteria as well as the presence or absence of related pathologies. Subsequently, and independently for each of the indexes, an algorithm was constructed by applying the conditions in Appendix 1 of Supplementary Material in order to catalog as appropriate or inappropriate each of these drugs, finally noting the number of inadequacies. We excluded from the Beers and STOPP criteria drugs not available in Spain. START criteria were not applied because patient follow-up is required. PIP associated with a history of falls and fractures were also omitted as only information at discharge was consulted, without follow-up data. We used the statistical package Stata version 10.0 for data analysis. ¿ Descriptive analysis of the study population characteristics: frequency distribution for the qualitative variables, and measures of central tendency and dispersion for the quantitative variables. ¿ Estimated prevalence of PIP for each criteria and its 95 % confidence interval (CI), globally and stratified for different categories of the study variables. ¿ Degree of agreement between the two criteria (Beers and STOPP) using the Kappa statistic. ¿ We estimated the strength of the association between each variable and the presence of at least one PIP by calculating the odds ratio (OR) and 95% confidence interval (CI). Adjustment for confounding variables was performed using a multiple logistics regression model that included all variables with a statistically significant effect, along with the patients¿ sex and age. Both saturated and selected models were estimated by a stepwise forward algorithm with an entry level of p < 0.25. A population of 624 patients, with somewhat more men than women, was included in our study. The median age was 78 years old (range 65¿95). The population was distributed evenly in three groups of services: Surgery, Internal Medicine and Other Medical Services, and 32.5 % of the sample suffered from high comorbidity (Charlson Comorbidity Index ¿ 4). The number of drugs prescribed presented a median value of eight (range 1¿21). Around 30% of all patients took 6 or less drugs and more than 20% of them took more than 11 drugs. The median length of stay in hospital was 7 days, ranging from 1 to 105 days. A total of 5350 medications were prescribed to the patients included in our study. The median number of medications per patient was 8 (range 1-21). Omeprazole was the most frequently prescribed drug (72.6% of patients), followed by furosemide (44.7% of patients) and aspirin (36.7% of patients). The overall frequency of PIP was 22.9 % according to the Beers criteria (95 % CI 19.6¿26.2 %) and 38.5 % (95 % CI 34.6¿42.3 %) according to the STOPP criteria; in 13.6 % of the patients the prescriptions were simultaneously inappropriate for Beers and STOPP criteria. Only 13 out of 143 patients with PIP showed more than one prescribed drug that met Beers criteria for inappropriateness; under STOPP criteria there were 64 patients (from 240) with more than one PIP. We estimated that the degree of agreement between the two criteria is 65.9 %, with Kappa 21.9 (SD 3.7) underlining a poor agreement between them. No differences were found in the proportion of inappropriateness by age, gender or length of hospital stay. Differences were, however, found in view of discharge service, the frequency being significantly higher among patients from Internal Medicine (33.33% vs. 19.92% in other medical services and 17.28% in Surgery according to STOPP; 50.28% vs. 35.55% in other medical services and 31.41% in Surgery according to Beers criteria). There was a significantly higher frequency of PIP when the Charlson Index was 4 or 5 compared to an index of 3 or less, but only under Beers criteria (p < 0.005). PIP frequency was seen to rise significantly with the number of drugs prescribed according to both criteria (p < 0.01). We found, for Beers criteria that the drugs most commonly resulting in inappropriate prescribing were (in this order): alpha blockers, NSAIDs, and calcium channel antagonists. For STOPP criteria, they were aspirin and NSAIDs. When stratified by sex, the most noteworthy difference is the predominance of calcium antagonists and digoxin in women and alpha blockers in men for Beers criteria, whereas for STOPP criteria they were NSAIDs in men and calcium channel antagonists and loop diuretics in women. Only five drugs were responsible for 81.1 and 76.7 % of PIP for Beers and STOPP criteria, respectively. The most frequent drugs that contributed to PIP using the Beers criteria were alpha-blockers (such as doxazosin, prazosin, and terazosin), which were considered to be potentially inappropriate in 65.7% of the prescriptions or 48 patients (when prescribed in the presence of syncope, hypertension, or urinary incontinence). While digoxin at a dose of >0.125 mg was inappropriate in all 19 prescriptions according to the Beers criteria, calcium antagonists were considered to be inappropriate in only 23 patients (44.2% of the prescriptions, which were those associated with heart failure or chronic constipation). When the STOPP criteria were used for PIP identification, the most common causes of inappropriateness were nonsteroidal anti-inflammatory drugs (NSAIDs) and aspirin, which were inappropriate in 65 (59.1%) and 104 (50.5%) of the prescriptions, respectively. NSAIDs are classified as inappropriate when prescribed in the presence of gastroduodenal ulceration without the concomitant use of antiulcer drugs, in the presence of digestive bleeding, or in patients with moderate to severe hypertension. The prescription of aspirin is potentially inappropriate in the presence of hemorrhagic disease or gastroduodenal ulceration, as for NSAIDs, as well as when associated with warfarin without the concomitant use of antiulcer drugs. Aspirin at a dose of >150 mg accounted for another 21 prescriptions, all of which were identified as inappropriate by the STOPP criteria. The patients¿ sex and length of hospital stay did not seem to be associated with PIP, unlike the type of hospital service, which was found to have a statistically significant influence (p < 0.01) according to both criteria. The risk of PIP was higher among patients discharged from the internal medicine service (Beers: aOR, 2.7; 95% CI, 1.4¿5.2; STOPP: aOR, 2.3; 95% CI, 1.3¿4.0) than among patients discharged from surgical services. The higher risk of PIP among patients discharged from the internal medicine service was not maintained when adjusting for confounding variables. Increasing age was a significant risk factor for PIP only when adjusting for other pathological conditions and the number of drugs using the Beers criteria. The Charlson comorbidity index was a risk factor for PIP according to the Beers criteria (aOR, 1.1; 95% CI, 1.0¿1.3), but not when PIP were defined by the STOPP criteria. The number of prescribed drugs was another evident risk factor for PIP. Each additional drug increased the risk of PIP by 14% (Beers criteria) or 15% (STOPP criteria). Specific pathological conditions that influenced the occurrence of PIP as identified by the Beers criteria were hypertension, cardiac failure, arrhythmia, permanent atrial fibrillation, renal failure, prostatism, and particularly dementia (aOR, 8.2; 95% CI, 4.0¿16.6). Conversely, preexisting ischemic cardiopathy, cerebrovascular disease, and peripheral arterial disease acted as protective factors (aOR, 0.4; 95% CI, 0.2¿0.7). Analysis of the remaining analyzed pathological conditions revealed no significant association. When the STOPP criteria were used for PIP identification, only hypertension, chronic obstructive pulmonary disease, and gout increased the risk of PIP, while the protective factors were identical to those identified using the Beers criteria (aOR, 0.3; 95% CI, 0.2¿0.5). Considering only the PIP associated with the five drugs or drugs groups most frequently used (Beers criteria: alpha-blockers, NSAIDs, calcium channel antagonists, digoxin, and benzodiazepines; STOPP criteria: aspirin, NSAIDs, loop diuretics, calcium channel antagonists and aspirin dose of >150 mg), the frequency of PIP was reduced from 22.9% to 19.7% and from 38.4% to 30.4%, respectively. The associated factors were almost identical between the two criteria (data not shown), the only difference being a lower effect of dementia in the STOPP criteria than in the Beers criteria (OR, 3.1; 95% CI, 1.5¿6.3). According to STOPP criteria, the frequency of hospital readmissions within 30 days after discharge was higher for patients without PIP (p = 0.053) and there was a significant increase in the average number of domiciliary visits (43.7 versus 29.7) in patients with PIP, almost significantly. For PIP identified by Beers criteria, no difference regarding health outcomes was observed between the two groups of patients generated. Patients discharged from the Internal Medicine service showed higher mortality rates than those discharged from Surgery. Besides, the comorbidities associated with increased mortality were arrhythmia and dementia. Neither STOPP, nor Beers criteria had a significant influence on mortality, nor did the intersection between the two types of criteria. For STOPP criteria, a significant and independent increase in mortality was observed when aspirin was inappropriately prescribed. This association was not found when all prescriptions of aspirin were analyzed. To conclude, the frequency of PIP at discharge is very high, as it affects about one in five patients with reference to the more restrictive criteria of Beers, or four out of 10 with STOPP criteria. Moreover, it increases with the number of prescribed drugs. This high prevalence of PIP at discharge should be a concern in Spanish hospitals. It is possible to identify a few drugs responsible for most cases of PIP, which means that more efficient surveillance and control procedures could reduce it. By intervening in five drug groups, about 80 % of PIP might be avoided according to either of the two criteria. The limited number of drugs involved in the majority of PIP suggests that simplifying STOPP and Beers criteria by reducing the number of covered items could facilitate their use without affecting usefulness. Pharmacists must be involved in PIP detection at hospital discharge; establishing an appropriate target population group and choosing a few drugs to review could be the most efficient approach. Risk factors for PIP according to both criteria used (Beers and STOPP) were discharge from the internal medicine service, a higher number of prescribed drugs, and the presence of moderate to severe hypertension. Conversely, preexisting ischemic cardiopathy, cerebrovascular disease, peripheral arterial disease, and arterial occlusion acted as protective factors. For the Beers criteria, additional pathological conditions were found to increase the risk of PIP, namely cardiac failure, arrhythmia, permanent atrial fibrillation, renal failure, and particularly dementia. Overall, the results of this study revealed differences between the risk factors for PIP depending on the type of criteria used to define PIP. However, both sets of criteria highlight the importance of polypharmacy and hypertension management on prevention of PIP. Our results do not confirm the existence of a relationship between potentially inappropriate prescriptions identified by Beers or STOPP criteria and the use of health services in the medium term, although a significantly higher number of visits was recorded. No association was found between the prescription of potentially inappropriate drugs and post-hospital mortality with any of the two sets of criteria used for their detection. However, the inappropriate prescription of aspirin according to the STOPP criteria behaved as an independent risk factor for mortality. We lacked information that would allow us to limit follow-up periods after discharge; furthermore, we had no information on the treatment changes performed by general practitioners. Future research needs to be done to determine the short-term effects of PIP at discharge on the use of health care resources or on mortality, as well as the influence of medications review and monitoring of prescriptions by general practitioners (family doctors), ideally in collaboration with clinical pharmacists.