Eur J Clin Pharmacol (1993) 45:199-203
l Q ecelleg @Springer-Verlag 1993
Originals Drug related hospital admissions Results from an intervention program
J. Hallas I, B. Harvald 2, J. W o r m 3, J. B e c k - N i e l s e n 3, L. E G r a m 1, E. G r o d u m 1, N. D a m s b o 4, J. Schou 5, H. K r o m a n n - A n d e r s e n 5, and F. Frolund 5
Department of Clinical Pharmacology, Institute of Medical Biology, Odense, Denmark 2Medical Department C, Odense University Hospital, Odense, Denmark 3Geriatric Department Q, Odense University Hospital, Odense, Denmark 4Department of General Medicine, ISH, Odense University Hospital, Odense, Denmark 5Danish Committee on Adverse Drug Reactions, National Health Board, Copenhagen, Denmark Received: October 21,1992/Accepted in revised form: March 9, 1993
Summary. As part of a high-intensity monitoring study of
drug events as the cause of admission to departments of internal medicine, the effect of an educational intervention p r o g r a m m e was studied. Two departments were included, one specialising in geriatrics and one that received patients by non-selected referral. The series consisted of 607 consecutive admissions studied before and 703 after the intervention. The drug events considered were adverse drug reactions and dose-related therapeutic failures, mainly due to non-compliance. A modest, statistically non-significant decrease in drug related hospital admissions ( D R H ) was seen, from 14 % before to 13 % after the intervention period. However, D R H s classified as definitely avoidable showed the significant decrease of 83 %. T h e r e was no apparent relationship between the topics selected for the intervention p r o g r a m m e and changes in the pattern of D R H s . No relationship between alterations in sales data and hospital admissions caused by a given drug could be demonstrated. A blinded external evaluation of case abstracts did not disclose any significant shift in the investigators' assessments. The intervention m a y have had an non-specific effect on avoidable D R H s . K e y words: Drug education, Hospital admission; adverse
drug reactions, drug utilisation, intervention
The contribution of adverse drug reactions and related events to admissions to departments of internal medicine has often b e e n investigated in the past three decades [13]. The results are often difficult to interpret in terms of risk in a strict epidemiological sense, as the appropriate "denominators", the size of the background population, its demographic characteristics, drug utilisation and the incidence of hospital admissions are only rarely available or reported [4]. Instead, the surveys m a y be interpreted in terms of quality assurance or health economics and so m a y be related to potential savings by reducing admission numbers. This has been an important aspect of some sur-
veys [5]. However, to our knowledge there has b e e n no published report of intervention p r o g r a m m e s aimed at reducing drug related hospital admissions in general. The aim of the present study was to evaluate the effect of an educational intervention on drug related hospital admissions (DRHs), by comparing admissions to two hospital departments before and after an intervention programme, targeted at the prescribing physicians. The study was conducted in tandem with a comprehensive survey to evaluate the role of drug related hospital admissions within different subspecialities of internal medicine [6-9]. Those studies had shown that, amongst the admissions to the six departments covering the entire field of internal medicine, in all about 10 % were drug related. A m o n g s t them, about one third (3-4 %) were judged in principle as avoidable. The avoidable hospital admissions would be the logical target of an intervention p r o g r a m m e . This report present the results of such an intervention and discuss the methodological issues entailed in this type of research.
Materials and m e t h o d s
Material During a period of 15 months (1988-1989) six departments of internal medicine at the Odense University Hospital in Denmark participated in a high-intensity monitoring programme of admissions caused by drug events. The series comprised 1999 admissions and showed a proportion of DRHs of 11.4 % of all admissions for a geographically defined population. In relation to this survey, a formal educational intervention programme was carried out, directed towards primary health care. The two departments first investigated were selected for reinvestigation. The remaining departments were investigated to some extent while the intervention programme was ongoing and the results from them have not been included in the analysis. Of the two reinvestigated departments, one primarily received its patients by non-selected referral (Department C), and one was subspecialised in geriatrics (Department Q). The pre-intervention series consisted of 321 and 286 consecutive admissions from Departments C and Q, respectively. The post-intervention series comprised 332 (C) and 371 (Q) consecutive admissions, collected during over a
200 period of 4 months, about 16-18 months after the first study period. Excluded from these figures were transferrals and planned admissions of patients not residing in the Odense area, 45 admissions in all. The rationale for this restriction has previously been discussed in detail [4]. The total number of admissions into both departments was 3007 in 1988 and 3184 in 1989.
citizens, were assumed to apply to Odense as well [10,11]. Sales were expressed by the "defined daily dose" methodology developed by the Nordic Council of Medicine to combine utilisation data for different drugs within the same therapeutic group [12].
Analysis Definitions and assessment The drug events considered were adverse drug reactions, defined as any unintended and undesirable effect of a drug, and dose-related therapeutic failures, defined as a lack of therapeutic effect that could be causally related either to drug non-compliance, too low a prescribed dose, interactions, recent dose reduction/discontinuation or inadequate therapeutic monitoring The evaluation of the cases followed a three-stage scheme: 1. Assessment of the certainty of a causal relationship between drug intake and the symptom suspected of being a drug event. The results presented here concern definite, probable or possible drug events unless otherwise specified. 2. The contribution of the symptom suspected of being a drug event to the causes of admission. All drug events presented were rated as the dominant or a contributory cause of admission. 3. The possibility of avoiding the drug event due to efforts by the health service personnel responsible for therapy. This evaluation of "avoidability" required reasonable certainty of a true drug event and was only carried out for 'definite' or 'probable' DRHs. "Avoidability" was scored on a rank scale, with 1,2 and 3 representing non-avoidable, possibly avoidable and definitely avoidable drug events. In brief, definitely avoidable drug events represent adverse events resulting from erratic prescribing, e. g. due to a lack of a definite indication, neglect of an overt contraindication or use of a clearly inadequate dose. Non-avoidable drug events were those resulting from well-indicated and well-managed therapy with reasonable weighing of risks and benefits. The data collection scheme and the criteria for evaluation have been described in detail elsewhere [6].
No changes in the referral policy of Odense University Hospital during the studies were produced by administrative action. However, for various reasons, the number of admissions to all medical departments of Odense University Hospital increased by 7.9 % from 1988 to 1989. There was no reason to assume that the number of D R H s would follow the same trend, since the size and age-distribution of the Odense population were virtually unchanged from 1988 to 1989, as was drug utilisation in general. Consequently, the analysis was carried out according to the basic presumption that an intervention effect should be expressed by a change in the absolute annual number of D R H s arising in the departments involved, and not by a change in the percentage of DRHs. The annual number of D R H s in a department in a given year was estimated by a direct standardisation technique, in which the characteristics of the sample studied from a given department were extrapolated to apply to the entire annual patient population in that department. Thus, estimates of annual numbers of DRHs, or subsets thereof, were performed on a compilation of the yearly patient populations for the two departments. Apart from expressing the intervention effect by the desired measure, i.e. absolute numbers of DRH, this standardisation also adjusts for possible arbitrary changes in the percentage of D R H s that could be produced, e. g. due to sampling relatively more geriatric patients after than before the intervention. The standardised figures are conservative when compared to crude percentages, i. e. more likely to show no change. Statistical testing of changes in the number of D R H was carried out by the Mantel-Haenzel test with each department forming a separate stratum. Ninety-five percent confidence intervals (CI) for rate ratios were calculated by the method described by Morris and Gardner [13]. Spearman's rank correlation analysis was used for comparison of "avoidability" assessments.
Intervention The intervention programme consisted of three elements: 1. The " D R H discharge summary", a letter sent to the patient's general practitioner by the investigator, summarising the D R H problem and stating the basis for the conclusion. These letters served several purposes. By informing the prescribing physician of this particular drug event, similar episodes in future should have been prevented. In addition, the communications might have had a nonspecific effect on alertness by informing the physician of the study. Finally, the D R H discharge summaries invited the family physician to comment on the D R H , which proved to be a valuable source of information [6]. 2. Eight monographs were distributed to all general practitioners within the region and to all interested medical registrars. They covered a selection of the drug events observed during the first phase of the study, illustrated by case histories, and supplemented by authoritative reviews of current therapeutic recommendations. 3. Four evening symposia with discussions of the observed drug events were held. All general practitioners and medical registrars within the region were invited.
Drug utilisation data Drug sales statistics were obtained to evaluate whether any link between changes in event rates and sales figures could be demonstrated for a given drug. Sales statistics could not be obtained specifically for the Odense region. Instead, the data for the Funen County, an area with 458,000 inhabitants, including the 205,000 Odense
Validation A shift in the investigators' causality assessment could produce a spurious change in recorded DRHs. This possibility was evaluated by a blinded, external rating of cases. Sixty randomly selected case abstracts containing the investigator's own D R H report, the hospital's discharge summary and the patient's general practitioner's comments on the D R H discharge summary, if any, were evaluated by an external peer group consisting of a pharmacologist (JS), a specialist in internal medicine (HKA) and a general practitioner (FF). The selected cases comprised 10 definite/probable DRHs, 10 other drug events and 10 cases without drug events, in all 30 cases from each of the study periods (pre- and postintervention). The peer group was blinded with regard to the period from which the case was collected, but used the same criteria for evaluation as the investigators.
Results A c o m p a r i s o n o f all p r e - a n d p o s t - i n t e r v e n t i o n a d m i s sions, c a r r i e d o u t s e p a r a t e l y f o r t h e t w o d e p a r t m e n t s d i d n o t r e v e a l a n y s i g n i f i c a n t d i f f e r e n c e in age, s e x o r in t h e n u m b e r o f d r u g s p r e s c r i b e d ( P > 0.35). Eighty-four DRHs were recorded before the intervent i o n a n d 88 a f t e r it (14 % a n d 13 % o f a d m i s s i o n s to t h e t w o d e p a r t m e n t s t o g e t h e r , P = 0.46). T h e e s t i m a t e d an-
201 Table 1. Pre- and post-intervention numbers of subjects with a given characteristic in the sample and estimated annual admission number into both departments Pre-intervention (n = 607) Observed number Definite + probable + possible DRHs - Adverse drug reactions Therapeutic failures Definite + probable DRHs Definitely avoidable DRHs - Possibly avoidable DRHs Non-avoidable DRHs Definitely + possibly avoidable DRHs - Therapy prescribed by general practitioners - Therapy prescribed by hospital doctors Others Hospitalized patients taking prescribed medicine
84 67 17 58 14 19 25 33 19 10
Post-intervention (n = 703) Estimated annual number 418 327 91 304 71 103 129 175 103 52
Observed number 88 72 16 58 3 22 33 25 13 10
Estimated annual number 383 316 67 268 12 102 151 114 58
Ratio (post/pre) of estimated annual numbers (CI)
0.92 (0.67 1.22) 0.97 (0.68-1.33) 0.74 (0.35-1.37) 0.88 (0.60-1.25) 0.17 (0.03~3.45)* 0.99 (0.51-l.75) 1.17 (0.68-1.91) 0.65 (0.37-1.05) 0.56 (0.26-1.06)
50
0.96 (0.36-2.0l)
4
20
2
6
0.30 (0.03-l.07)
533
2593
593
2655
1.02 (0.91 1.13)
DRH, Drug related hospital admission; CI, 95 % confidence intervai *P <0.01 nual n u m b e r s of D R H s d e c r e a s e d f r o m 418 to 383, i. e. 35 admissions. T h e rate ratio was 0.92 (C 0.67-1.22). A sub-analysis of avoidable D R H s (Table 1) s h o w e d a substantial, statistically significant d e c r e a s e in definitely avoidable D R H s . F o u r t e e n cases were r e c o r d e d b e f o r e and t h r e e after the intervention (rate ratio 0.17, CI 0.030.45). F o r possibly avoidable D R H the estimated annual n u m b e r s of pre- and p o s t - i n t e r v e n t i o n D R H s were 103 and 102 (rate ratio 0.99, NS). A s shown in Table 1, the decrease in possible - and definitely avoidable D R H s app e a r e d to be m o s t p r o n o u n c e d a m o n g patients w h o s e t h e r a p y was p r e s c r i b e d by general practitioners, although this t e n d e n c y was not statistically significant. T h e type of D R H influenced by the intervention showed no obvious relationship to the topics focused on during the intervention. As s h o w n in Table 2, the p a t t e r n of D R H did not change significantly. R e g i o n a l drug c o n s u m p t i o n was fairly stable during the two data collection periods, and a consistent link b e t w e e n changes in the utilisation of specific drugs and the drug-related hospital admissions was not seen (Table 3). Confidence intervals for changes in D R H rates were wide, but the point estimates almost invariably s h o w e d changes c o n t r a r y to those of the changes in drug utilisation.
Validation T h e results of the external evaluation are shown in Table 4. T h e external p e e r group t e n d e d to classify m o r e cases as " n o drug events" than the investigators, but t h e r e was no significant difference in their rating of cases f r o m the pre- to the p o s t - i n t e r v e n t i o n periods. A m o n g the 6 cases f r o m the p r e - i n t e r v e n t i o n period which b o t h groups had classified as definite/probable D R H s , the s u m m e d score of avoidability was 12 for the p e e r g r o u p and 12 for
the investigators ( S p e a r m a n ' s r a n k correlation = 0.56, adjusted for ties). T h e c o r r e s p o n d i n g figures for 9 definite/probable D R H s f r o m the p o s t - i n t e r v e n t i o n p e r i o d was 18 for the p e e r group and 12 for the investigators ( S p e a r m a n ' s rank correlation = 0.60, adjusted for ties).
D i s c u s s i o n
A n u m b e r of studies have successfully shown the beneficial effect intervention by pharmacists or clinical p h a r m a cologists on prescribing quality, with prescription p a r a m e ters as the o u t c o m e [14-18]. H o w e v e r , there have b e e n few studies using a clinical o u t c o m e as the end-point [19]. Successful intervention has b e e n r e p o r t e d for drug-induced h y p o k a l e m i a [20], p o o r h y p e r t e n s i o n control [21] and fatal d e x t r o p r o p o x y p h e n e poisoning [22]. To our Table 2. Number of drug events causing hospital admission in preand post-intervention surveys. Figures in parentheses indicate number of cases classified as definitely or probably avoidable
Adverse drug reactions Analgesics andanti-rheumatics" Psychotropicdrugs Cardiovascular drugs" Anti-diabetics Miscellaneous Dose related therapeutic failures Non-compliance a Others All drug related hospital admissions "Subject for intervention
Pre-intervention (n - 607) Drug events 67 (26) 20 (5) 16 (6) 12 (5) 10 (4) 9 (6) 17 (7) 11 (4) 6 (3) 84 (33)
Post-intervention (n = 703) Drug events 72 (20) 17 (4) 25 (7) 15 (3) 2 (1) 13 (5) 16 (5) 8 (2) 8 (3) 88 (25)
202 3. Regional sale of drugs involved in drug related hospital admission (DRH). Sales are expressed in defined daily doses (DDD) per 1,000 inhabitants per day Table
Drug
Sale 1988
Ratio of sale 1989/1988
NSAID Analgesics Psychotropic drugs Cardiovascular drugs - Diuretics - Digoxin
20.2 83.9 124.3 190.0 137.3 8.0
1.06 1.03 0.88 0.97 0.95 1.15
Estimated DRH rate ratio 1989/1988 (CI) 0.86 (0.25-2.1) 0.83 (0.33-1.7) 1.33 (0.68-2.4) 1.46 (0.64-2.9) 2.25 (0.70-5.8) 0.00 NA
5.8 8.5
1.00 1.22
0.21 (0.02q3.7)* 2.70 (0.51-9.5)
Anti-diabetics Corticosteroids
CI, 95 % confidence interval; NA, CI method not applicable * P < 0.01 4. External peer group's evaluation of cases assessed by the study group as definite/probable DRH (10 cases), other drug events (10 cases) or no drug events (10 cases), with 3 x 10 cases from each period (pre-/post-intervention). External peer group was blind to what period the cases belonged to Table
Classification Definite/probableDRHs Other drug events No drug event
Pre-intervention 8 4 18
Post-intervention 13 3 14
DRH, Drug related hospital admission Z 2= 2.26; P = 0.32
knowledge, no study has investigated the impact of an intervention p r o g r a m m e on D R H s in general. Generally, only a modest, statistically non-significant decrease in the D R H s was seen. In a substantial part of the pre-intervention D R H s an intervention effect was not expected, since they were "not avoidable" D R H s in nature. These D R H s occurred in a therapy in full accordance with recommendations, weighing the benefits and risks [6]. The substantial change in avoidable D R H s strongly suggested a true intervention effect. This decrease seemed to be most p r o n o u n c e d in the group for which the intervention primarily had b e e n targeted, i.e. for general practitioners. We could not demonstrate a clear consistency between the topics selected for intervention and changes in the pattern of D R H s , avoidable or not. This suggests that the intervention effect is largely non-specific, e.g. produced by an increased general alertness a m o n g drug prescribers, rather than by a direct use of the distributed material. We know of no other formal intervention or major therapeutic advances that could explain the decrease in avoidable D R H s . As suggested in Table 3, a reduction in the incidence of serious adverse reactions need not be accompanied by a similar decrease in over-all drug sale. Some D R H surveys indicate that drugs generally are consumed in very large quantities for each serious adverse reaction elicited [4, 23]. The amount of drugs used in an inappropriate and hazardous m a n n e r is usually only a minor proportion of the over-all utilisation [14]. Consequently, the eradication of inappropriate prescribing of a drug m a y have a great
impact on the n u m b e r of adverse reactions but little effect on the drug use, which will largely be determined by other, more powerful factors. As an example, a recent campaign from D e n m a r k m a n a g e d to reduce fatal dextropropoxyphene intoxications by 45 %, while the dextropropoxyphene sale decreased by only 22 % [22]. The design of the present study was that of an uncontrolled experiment. There was a recording of hospital admission data before and after an intervention, but no equivalent recording from a hospital where the prescribing physicians had not been exposed to the intervention programme. Having a control group poses several serious problems. Firstly, it would have required involvement of an other region as control. Secondly, the registration of D R H s would in itself have had a strong intervention effect in the control region. Finally, to avoid bias in the registration of D R H cases, the data collection and evaluation of cases should be carried out blindly by the investigators. This was considered highly unfeasible. Instead, we endeavored to use the same data collection techniques and to confirm the results by a blinded external evaluation, if possible. In line with a similar assessment in an earlier part of the study p r o g r a m m e [4], there was no clear indication that a shift in the causality rating would explain the recorded change in avoidable D R H s . The differences actually observed between external and investigator assessment may reflect that the commonly used algorithms for causality assessment do not work well in a setting like the present and show a very large inter-observer variability [24]. Although the intervention study involved registration of m o r e than 1300 hospital admissions, the size of the study can still be considered too small to obtain precise estimates of the consequences of the intervention study. The two departments studied are probably representative of medical departments in general. D e p a r t m e n t C, which had the highest admission numbers in this material, received the majority of its patients by acute, non-selected referral. The average in-patient time was 13 days. With an average daily cost of 400 US $ per bed, the estimated savings of 35 admissions in one year in the two departments studied (180,000 US $) could have financed the whole D R H project in all six departments for all three years. If the results from this survey can be generalised, a monitoring of the consequences of out-patient drug use is not only a potential source of knowledge, but also has a very favorable cost-benefit profile.
References
1. Seidl LG, Thornton GE, Smith JW, Cluff LE (1966) Studies on the epidemiology of adverse drug reactions. Bull John Hopkins Hosp 119:299-315 2. Miller RR (1974) Boston Collaborative Drug Surveillance Program. Hospital admissions due to adverse drug reactions. Arch Int Med 134:219-222 3. Bergman U, Wiholm B-E (1981) Drug related problems causing admission to a medical clinic. Eur J Clin Pharmaco120 193-200 4. Hallas J, Gram LF, Grodum E, Damsbo N, BrCsen K, Haghfelt T, Harvald B, Beck-Nielsen J, Worm J, Jensen KB, Davidsen O, Frandsen NE, Hagen C, Andersen M, FrOlund F, Kromann-An-
203 dersen H, Schou J (1992) Drug related admissions to medical wards. A population based survey. Br J Clin Pharmaco133:61~68 5. Italian Group on Intensive Care Evaluation (1987) Epidemiology of adverse drug reactions in intensive care units. A multicentre prospective survey. Eur J Clin Pharmaco131:507-512 6. Hallas J, Harvald B, Gram LF, Grodum E, BrCsen K, Haghfelt T, Damsbo N (1990) Drug related hospitalisations. The role of definitions and intensity of the data collection, and the possibilityof prevention. J Intern Med 228:83-90 7. Hallas J, Haghfelt T, Gram LF, Grodum E, Damsbo N (1990) Drug related admissions to a cardiology department; frequency and avoidability. J Intern Med 228:379-384 8. Hallas J, Jensen KB, Grodum E, Damsbo N, Gram LF (1991) Drug related admissions to a department of medical gastroenterology. The role of self-medicated and prescribed medicine. Scand J Gastroentero126: 174-180. 9. Hallas J, Worm J, Beck-Nielsen J, Gram LF, Grodum E, Damsbo N, BrCsen K (1991) Drug related events and drug utilisation in patients admitted to a geriatric hospital department. Dan Med Bull 38:417-420 10. Anonymous. (1989) Drug comsumption in Denmark 1988. D ansk La:gemiddelstatistik 1989. Strcdamvej 50 B, DK-2100 O, Denmark 11. Anonymous. (1990) Drug comsumption in Denmark 1989. Dansk Laegemiddelstatistik 1990. Streidamvej 50B, DK-2100 O, Denmark 12. Nordic Council of Medicine (1985) Nordic drug index with defined daily doses. Nordic Statistics on Medicine 1981-1983, part 2 13. Morris JA, Gardner MJ (1989) Calculating confidence intervals for relative risks, odds ratios, and standardised ratios and rates. In: Gardner JA, Altman DG (eds) Statistics with confidence. BMJ, edition, London 1989:50~53 14. Hawkey CJ, Hodgson S, Norman A, Daneshmend TK, Garner ST (1990) Effect of reactive pharmacy- intervention on quality of hospital prescribing. BMJ 300:986-990
15. Ray WA, Blazer DG, Shaffner W, Federspeil CF, Fink R (1986) Reducing the long-term diazepam prescribing in office practice. A controlled trial of educational visits. JAMA 256:2536-2539 16. Owens N J, Sherburne N J, Silliman RA, Fretwell MD (1990) The senior care study. The optimal use of medications in acutely ill older patients. J Am Geriatr Soc 38:1082-1087 17. Buck CR, White KL (1974) Peer review, impact of a system based on billing claims. N Engl J Med 291 : 877-983 18. Diwan M Angunawela I, Tomson G (1989) A method to evaluate effect of drug information, randomisation in groups: an experience from Sri Lanka. Eur J Clin Pharmaco136 [Suppl]: 102 A 19. Gurwitz JH, Soumerai SB, Avorn J (1990) Improving medication prescribing and utilisation in the nursing home. J Am Geriatr Soc 38:542 552 20. FitzGerald GR, Delaney E, Cushen M, FitzGerald MG (1989) Diuretic associated hopykalemia in hospital admission. Res Clin Forums 11:49-53 21. Williamson DH, Cooper JW, Kotzan JA (1984) Consultant pharmacist intervention on antihypertensive therapy in a geriatric long-term facility. Hosp Formu119:123-128 22. Ott R Hansen PB, Dalhoff KR Loft SH, Poulsen HE (1990) Consumption of overdose and fatal poisoning with analgesics in Denmark 1979-86 (in Danish with an English abstract). Ugeskr Laeger 152:250-252 23. Jick H (1974) Drugs - remarkably nontoxic. N Engl J Med 291: 824-828 24. Louik C, Lacoutre PG, Michell AA, Kauffman R, Lovejoy FH, Yaffe SJ. Shapiro S (1985) A study of adverse reaction algorithms in a drug surveillance program. Clin Pharmacol Ther 38:183-187 J. Hallas, MD Department of Clinical Pharmacology Institute of Medical Biology Winsl~lwsparken 19,3 DK-5000 Odense C Denmark