Although the pharmacists at the hospital were notified when patients were assigned to the control group, no information exchange occurred for these patients; rather, care was provided as usual. (Approximately three weeks after the study began, a new computer system was installed, and pharmacists began documenting interventions for all patients. Documentation would not have occurred routinely for control patients during the first three weeks if their enrollment had not been disclosed to the pharmacists.)

Patients were enrolled between January 3, 1996, and June 30, 1996. Pharmacy C was not added until May 7, 1996 (in an attempt to increase enrollment). All patients 18 years of age or older were eligible for enrollment. Patients were enrolled only once, regardless of how many hospital admissions they had during the enrollment period. The following types of patients were excluded from enrollment: (1) patients being treated for HIV- related illness, (2) patients admitted for drug or alcohol detoxification, (3) patients admitted for treatment of mental illness, and (4) patients admitted to the obstetrics unit for delivery or other obstetrical care. Pharmacists participating in the information-exchange program were not explicitly blinded to the study's objectives.

The project was approved by a university human subjects committee and by an institutional review board serving area hospitals.

Methods of System Evaluation

Patient identifiers were removed from data before it was provided to the research team, and the data abstractor was blinded with respect to study group. Data analyses were primarily done for two periods: the period when patients were hospitalized and the period from discharge to 60 days after discharge. Demographic variables analyzed included age, sex, source pharmacy, length of hospitalization, admitting (primary) diagnoses, number of chronic diseases, and number of admission medications.

All data were collected from existing clinical documentation. Pharmacists' notes (interventions), hospital computer system data (demographic variables), and physician documentation (diagnoses, chronic diseases) were collected. Interventions documented by pharmacists in the hospital and posthospital periods were counted as such if a specific drug therapy problem or potential problem and an action to resolve the problem were documented.

Results presented for the experimental and control groups include all patients randomly assigned to the respective groups (e.g., patients assigned to the experimental group for whom only partial information exchange occurred were still considered part of the experimental group).

Statistical analysis

Data were analyzed with Statistical Analysis System version 6.01 (SAS, Cary, NC). All statistical tests were two tailed, and the a priori level of significance was 0.05. However, because of the small sample size, analyses yielding higher p values (0.06-0.20) were considered for inclusion in models and further analyses. An initial univariate analysis was done to examine the impact of potentially confounding and interacting variables. For the categorical variable source pharmacy, Fisher's exact test was used because of the small number of patients originating from each of the four pharmacies; Pearson's chi-square test was used to compare data for nominal categorical variables, and the Mantel Haenszel chi-square test was used to compare data for ordinal categorical variables. The Cochran-Mantel Haenszel summary test was used to compare variables after controlling for confounding variables such as sex and age and was used for other analyses of stratified data. The Mantel-Haenszel estimate of the adjusted relative risk was used to assess the magnitude of association between dichotomous categorical variables.¹³ The Breslow-Day test for homogeneity was used to assess the appropriateness of summary tests for stratified variables. Analysis of variance and analysis of covariance were used to assess continuous variables. Group means were compared by using Tukey's test.

Results

A total of 47 patients from the ambulatory care pharmacies returned to their pharmacy for at least one prescription sometime during the 60-day postdischarge period. This represented 50% of all ambulatory care study patients discharged alive. More control group patients (58%) than experimental group patients (41%) returned to have at least one prescription filled (p = 0.05, chi-square test).

Information transfer

Interventions during hospitalization

With respect to the 96 medication-related in-hospital interventions, 43% (34 of 80) of all experimental group interventions were of the medication-addition type, compared with 19% (3 of 16) of all control group interventions (p = 0.08). Nineteen percent (15 of 80) of experimental group interventions and 31% (5 of 16) of control group medication-related interventions involved changing a dosage (p = 0.25). Twenty percent (16 of 80) of experimental group medication-related interventions involved a drug allergy, versus 6% (1 of 16) of control group interventions (p = 0.20). A significantly smaller proportion of medication-related experimental group interventions (19%, 15 of 80) than control group interventions (44%, 7 of 16) were of the remaining types listed in the preceding paragraph (p = 0.04). Overall, 46% (37 of 80) experimental group medication-related interventions were associated with maintenance drugs (e.g., antihypertensives, antiarrhythmics), compared with 50% (8 of 16) of medication-related interventions in the control group (p = 0.79).

Posthospital interventions

Compared with the control group, a significantly higher proportion of experimental group posthospital interventions were of the follow-up type (p = 0.02) and allergy- related type (p = 0.02). A significantly higher proportion of control group than experimental group interventions involved a request for a diagnosis (p = 0.02) and a request to discontinue a medication (p = 0.02). Although a larger proportion of control group than experimental group interventions involved a dosage change, the difference was not significant.

Of the posthospital interventions that included documentation of specific, identifiable drugs (51 of 57), 51% (19 of 37) of experimetal group interventions were associated with maintenance medications, versus 36% (5 of 14) for the control group; the difference was not significant.

Discussion

The project reported here is the first to formally evaluate the effectiveness of a system for exchanging patient information among pharmacists in different practice settings. Although some evidence of implementation of these systems is beginning to appear in the literature,^14,15 the contribution of these systems to patient care has not been known.

It appears that the information-exchange system was successful. In all five instances in which an ambulatory care pharmacy did not send all elements of the requested data, the ambulatory care pharmacist noted receipt of the request and documented the reason for not providing the information in each case, that the pharmacist had no medication information available for the patient. Thus, even in these five instances, the communication link between the hospital and the ambulatory care pharmacy was successful.

Sending of information from the hospital pharmacy was somewhat less successful, occurring 81% of the time. Although the reasons for failure to send this information were not formally documented, the hospital pharmacists indicated that a common reason was very late notification of a pharmacist that a patient was about to be discharged. Even though the hospital pharmacists were expected to send the information in this situation, they may have decided that it would be of limited use because of the time since discharge. Other reasons cited for failure of hospital pharmacists to send information included problems with the fax machine and pharmacy staffing shortages. The overall rate of complete information transfer was high (75%).

The method of identifying patients for the study involved manual comparison of hospital daily admission records with a list of patients from the source pharmacies. This approach was necessary because of the pilot nature of the study. If the information- exchange system were to be implemented beyond a pilot phase, a more automated approach would be warranted. For example, patients being hospitalized might be asked to designate their primary pharmacy when they fill out other admission forms. If the pharmacy participates in the information-transfer system, the admissions personnel could immediately fax the request for data.

If such a system is to improve patient outcomes, it seems necessary to first establish that it increases pharmacists' ability to intervene on behalf of patients. Given that the system was implemented successfully (i.e., the information was usually transferred), what effect did it have on pharmacist interventions? Experimental group patients were significantly more likely than the control group to have at least one in-hospital intervention. Furthermore, interventions for the experimental group appeared more likely to be of the type that would more commonly be identified with access to pre-admission patient histories (as was the case for experimental group patients), although not all the differences were significant. For example, identification of a drug missing from a regimen (which would potentially lead to its addition) and identification of an apparent drug allergy occurred slightly more frequently for experimental group patients. It is common for hospitals to omit information about medications patients were taking before they were hospitalized; Beers et al.3 found that 83% of patients had either an error of omission or an inappropriate inclusion in the admission medication history, with omission being the more common of the two.

In contrast, interventions for control group patients were significantly more likely to be of types that should be readily identified from access to just the inpatient medication list (and not pre-admission information), such as problems with the route or dosage form, drug interactions, duplicate therapy, or need for clarification of medication orders.

It is likely that pharmacists, knowing which patients were in the experimental group, focused more intensely on these patients and that this resulted in more in-hospital interventions for them. Indeed, a major goal was to focus pharmacists' efforts on recognizing potential problems in the drug regimen on the basis of knowledge of the patient's medication history. Ideally, this would contribute to the pharmacist's ability to identify himself or herself as that patient's caregiver and to begin to accept responsibility for the patient's care.

It is also possible that pharmacists, knowing they were under scrutiny, reacted by documenting more interventions for experimental group patients than they otherwise would have. It was not possible to separate the contribution of the receipt of information from that of the study environment in motivating the pharmacists to identify problems and document interventions. Furthermore, because it was not possible to blind pharmacists to the experimental group status of patients, the possibility of "discriminatory documentation" (i.e., an intentional effort on the part of the pharmacist to vary documentation practices according to the study group status of the patient) cannot be ruled out. Discriminatory documentation would most likely have resulted in fewer interventions being documented for control group patients. However, intervention types appeared to differ between the experimental and control groups and were consistent with expected uses of preadmission medication information, so it appears that discriminatory documentation did not play a major role.

The types of interventions most frequently documented for experimental group patients during hospitalization imply that pharmacists had received patient information that they otherwise would not have had. It is possible that this information could have come from somewhere other than the ambulatory care or LTC pharmacies; however, one would then expect that interventions for control group patients would be similar in number and type to those for experimental group patients. Instead, there were considerably fewer interventions for control patients, and those that occurred tended to be different.

Although a larger proportion of experimental group patients than control patients had at least one documented intervention after discharge, the difference was not significant. The LTC pharmacy documented interventions more often for control group patients than for experimental group patients (although the difference was not significant). In contrast, the ambulatory care pharmacies did not document any interventions for control patients. Because the LTC pharmacy did not routinely document day-to-day interventions (although these were routinely performed), data on LTC interventions came from monthly drug regimen review documentation. This meant that interventions performed by the LTC pharmacists early in the postdischarge period (such as identifying medications omitted from a patient's regimen, or resolving other more immediate problems) were not represented. For this reason, the LTC pharmacy interventions reported here may not represent the true number and distribution of interventions actually performed. Furthermore, because LTC pharmacists were routinely supplied patient information by the LTC facilities, it is less likely that the study's information- exchange system provided new information to these pharmacists to the same extent that it did to the ambulatory care and hospital pharmacists.

Use of several hospital sites rather than just one would have increased patient enrollment and allowed for subgroup analysis by hospital site. Consistent findings across hospitals may have offered further assurance that the findings were indeed a result of the information- exchange system, and inconsistencies may have enabled discussion about how site- related differences (such as staffing and motivation) influenced the results. However, no other hospitals in the area were available to participate in this study.

Further studies are warranted to establish whether information exchange has a favorable effect on patient outcomes. These studies should enroll more providers and more patients and target patients at higher risk for drug-related problems in order to have adequate statistical power and to more closely examine the impact of provider and patient characteristics on the effectiveness of information exchange.

Conclusion

A system of information exchange among pharmacists in different practice settings was implemented. Hospital and ambulatory care pharmacists documented more interventions for patients about whom information had been supplied than for patients for whom that information had not been supplied. No difference in intervention rates was observed for LTC pharmacists, who were already being supplied information by the LTC facilities about patients discharged from the hospital.

References

1. Rogers J, Curtis P. The concept and measurement of continuity in primary care. Am J Public Health 1980; 70:122-7.
2. Hennen BK. Continuity of care in family practice: dimensions of continuity. J Fam Pract 1975;2:371-2.
3. Beers MH, Dang J, Hasegawa J et al. Influence of hospitalization on drug therapy in the elderly. J Am Geriatr Soc 1989;37:679-83.
4. Rupp MT, Schondelmeyer SW, Wilson GT et al. Documenting prescribing errors and pharmacist interventions in community pharmacy practice. Am Pharm 1988;NS28(9):30-6.
5. Rupp MT. Value of community pharmacists' interventions to correct prescribing errors. Ann Pharmacother 1992;26:1580-4.
6. Dobie RL, Rascati KL. Documenting the value of pharmacist interventions. Am Pharm 1994;NS34(5):50-4.
7. Kimberlin CL, Berardo DH, Pendergast JF et al. Effects of an education program for community pharmacists on detecting drug-related problems in elderly patients. Med Care 1993; 31:451-68.
8. Currie JD, Chrischilles EA, Kuehl AK et al. Effect of a training program on community pharmacists' detection of and intervention in drug-related problems. J Am Pharm Assoc 1997; NS37:182-91.
9. Lipton HL, Bird JA. The impact of clinical pharmacists' consultation on geriatric patients' compliance and medical care use: a randomized controlled trial. Gerontologist 1994;34:307-15.
10. Lipton HL, Byrns PJ, Soumerai SB et al. Pharmacists as agents of change for rational drug therapy. Int J Technol Assess Health Care 1995;11:485-508.
11. Johnson JA, Bootman L. Drug-
related morbidity and mortality. A cost-of-illness model. Arch Intern Med 1995;155:1949-56.
12. Manasse HR Jr. Medication use in an imperfect world: drug misadventuring as an issue of public policy, part 1. Am J Hosp Pharm 1989;46:929-44.
13. Breslow NE, Day NE. Statistical methods in cancer research, volume II: the design and analysis of cohort studies. London: Oxford University Press; 1989.
14. Cameron B. The impact of pharmacy discharge planning on continuity of care. Can J Hosp Pharm 1994;47:101-9.
15. Foisy MM, Tseng A, Blaikie N. Pharmacists' provision of continuity of care to patients with human immunodeficiency virus infection. Am J Health-Syst Pharm 1996; 53:1013-7.

Address for Reprints: Dr. Angela K. Kuehl, 2949 Steindler Bldg. Dept. of Preventive Medicine, University of Iowa, Iowa City, IA 52242.