1 Mayo Clin Proc, July 2001, Vol 76 Alcohol Withdrawal in Inpatients 695 Original Article Symptom-Triggered Therapy for Alcohol Withdrawal Syndrome in Medical Inpatients THOMAS M. JAEGER, MD; ROBERT H. LOHR, MD; AND V. SHANE PANKRATZ, PHD Objective: To assess the efficacy of symptom-triggered therapy vs usual care for alcohol withdrawal syndrome (AWS) in medical inpatients. Patients and Methods: This study was a retrospective analysis of patients admitted to general medical services between January 1, 1995, and December 31, 1998, who experienced AWS during the admission. This study was conducted at Saint Marys Hospital, Rochester, Minn. Patients were identified from hospital discharge diagnoses and pharmacy data. Symptom-triggered therapy for AWS was initiated in Patients were divided into preimplementation ( ) and postimplementation ( ) cohorts. Age, sex, medical comorbid conditions, previous AWS (including seizures and delirium tremens), duration of treatment for AWS, benzodiazepine use and dose, complications of AWS, and adverse outcomes of treatment during the incident admission were abstracted from the medical records of eligible patients. Comorbid conditions were classified according to the Charlson comorbidity index. Differences between the cohorts were assessed with use of logistic regression models and analysis of covariance. Results: Review of medical records from 638 admissions (536 patients) yielded 216 admissions eligible for this study. After adjustment for age, sex, Charlson comorbidity index, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens, we found no significant difference between cohorts for duration of treatment (P=.16), benzodiazepine use (P=.21), total dose of benzodiazepine (P=.38), or total complication rate (P=.053). We did observe a significant difference in the occurrence of delirium tremens between the 2 treatment groups (P=.04). This was especially apparent for patients with no history of delirium tremens. Conclusions: Symptom-triggered therapy is effective treatment for AWS in medical inpatients. In this retrospective study, it did not result in shorter duration of treatment but was associated with a decreased occurrence of delirium tremens, the most severe and life-threatening complication of AWS. This result was most apparent in patients with no history of delirium tremens. Mayo Clin Proc. 2001;76: ANCOVA = analysis of covariance; AWS = alcohol withdrawal syndrome Alcohol withdrawal syndrome (AWS) is treated in many inpatient settings other than chemical dependency units. There are no reliable prognostic indices to estimate the occurrence or progression of AWS. Thus, treatment has been empiric and centered on substitution of other agents for alcohol, gradually tapering the dose of the agent used. 1,2 In the United States, benzodiazepines are used frequently for this purpose, and they are the only agents proven by placebo-controlled trials to prevent progression to serious complications of AWS. 1,3-5 The size and frequency of benzodiazepine dosage have usually been physician directed, but these drugs appear to be prescribed without regard to the severity of symptoms. A typical strategy that establishes benzodiazepine dosage only at From the Division of Community Internal Medicine (T.M.J.), Division of Area General Internal Medicine (R.H.L.), and Section of Biostatistics (V.S.P.), Mayo Clinic, Rochester, Minn. Address reprint requests and correspondence to Robert H. Lohr, MD, Division of Area General Internal Medicine, Mayo Clinic, 200 First St SW, Rochester, MN ( initial evaluation undertreats some patients and overmedicates others. Recent recommendations for treatment of AWS suggest a symptom-triggered approach based on frequent objective assessment of the patient. 1,6 Symptom-triggered therapy addresses the potential for under- or overmedicating by assessing symptoms on a real-time basis and then matching benzodiazepine dosage to symptom severity. Studies emanating from chemical dependency units have shown that benzodiazepines can be administered successfully when patients are symptomatic and withheld or decreased when they are less symptomatic. 6 Symptom-triggered therapy is a logical response to the variable symptoms and severity of AWS. In 1997, we instituted a program for the treatment of AWS for patients admitted to the general medical services at Saint Marys Hospital in Rochester, Minn. This program objectively assesses patient symptoms by using the Clinical Institute Withdrawal Assessment-Alcohol (revised) (CIWA-Ar), 7 and dosage of benzodiazepine is based on those assessments. Mayo Clin Proc. 2001;76: Mayo Foundation for Medical Education and Research
2 696 Alcohol Withdrawal in Inpatients Mayo Clin Proc, July 2001, Vol 76 The purpose of this study was to test the hypothesis that benzodiazepine therapy for AWS via a symptom-triggered strategy in patients admitted to general medical wards resulted in shorter duration of therapy and less total drug utilization compared with historical controls, as has been demonstrated in patients treated in chemical dependency units. 6 This study also compared complication rates and adverse outcomes of treatment associated with symptom-triggered administration of benzodiazepines vs usual care. PATIENTS AND METHODS Patients We retrospectively identified inpatients at Saint Marys Hospital treated for AWS from January 1, 1995, through December 31, 1998, by 2 methods. Hospital discharge diagnoses during the study period were reviewed for alcoholism, alcohol withdrawal, delirium tremens, and alcohol withdrawal seizures. In addition, hospital pharmacy data were reviewed to identify patients who received thiamine and benzodiazepines simultaneously during a given admission for the same period of time. Patients who did not provide written permission to be included in medical research studies at the Mayo Clinic were excluded as prescribed by Minnesota statute. The medical records from 536 patients were reviewed to obtain data on 638 admissions for the presence of alcohol withdrawal: 192 (30.1%) in 1995, 114 (17.9%) in 1996, 127 (19.9%) in 1997, and 205 (32.1%) in Of the 638 admissions reviewed, 216 (33.9%) were eligible for inclusion in the study: 51 (23.6%) in 1995, 33 (15.3%) in 1996, 52 (24.1%) in 1997, and 80 (37.0%) in Only patients cared for on general medical wards (Community Internal Medicine and General Internal Medicine) were analyzed. Patients who presented with an AWS seizure were not analyzed, as aggressive benzodiazepine treatment was generally initiated immediately in these patients. Identified patients were divided into a control group (preimplementation, ) and a treatment group (postimplementation, ). Patients in the preimplementation group were treated for AWS with usual care, ie, empiric benzodiazepine dosage usually on a tapering fixed-dose regimen or with as-needed doses at the discretion of medical staff but without a uniform pattern. Patients in the postimplementation group were treated according to a symptom-triggered protocol. After the medical team established the diagnosis of AWS and the appropriateness of symptom-triggered treatment, the CIWA-Ar protocol was initiated. Nursing and/or medical staff determined patient CIWA scores at 1- to 2-hour intervals during the time the patients were on the protocol. Dosage of benzodiazepine was determined by nursing staff, based on the patient s CIWA score and response to previous doses. If the patient s CIWA score was higher than 10, a typical starting dose of chlordiazepoxide was 50 to 100 mg with similar repeat doses until the CIWA score began to decline. The CIWA-Ar protocol and symptom-triggered therapy were introduced each year through didactic sessions to all postgraduate year 1 residents rotating on the general medical services. The nursing service was trained more extensively in use of the CIWA-Ar. Each registered nurse completed a 2-hour training session that included detailed instruction on the components of the CIWA-Ar as well as video demonstration of a simulated patient representing the range of symptom severity for each scale. Registered nurses were required to pass a posttest to verify competency in use of the protocol. All new registered nurses hired during the study period also completed the same training. Statistical Methods Age at admission, sex, medical comorbidities, prior AWS and/or AWS complications, current duration of treatment for alcohol withdrawal, benzodiazepine use, benzodiazepine dose, and current complications of alcohol withdrawal as well as adverse outcomes of treatment, including delirium tremens, seizures, leaving the hospital against medical advice, readmission for alcohol withdrawal within 7 days, transfer to the intensive care unit, and death, were abstracted from the medical records of eligible patients. Medical comorbidities were summarized in a single measure with use of the Charlson comorbidity index. 8 This weighted index takes into account the number and seriousness of comorbid conditions; the developers of the scale found that higher scores were associated with higher 1-year mortality. Patient and alcohol withdrawal treatment characteristics were summarized with means, SDs, medians, and ranges or frequencies and percentages. Duration of treatment for alcohol withdrawal for those patients treated with benzodiazepines was defined as the number of hours between the date and time of treatment initiation (first administration of benzodiazepine) and the date and time of treatment termination (last administration of benzodiazepine). For those patients not treated pharmacologically (ie, observed only), duration of treatment could not be determined. The length of hospitalization for patients admitted and discharged on the same day was set to 0.5 day. To facilitate statistical analysis, benzodiazepine dose was defined as the total oral lorazepam dosage equivalents (in milligrams) received during the duration of treatment. The benzodiazepine drugs administered and their dosage equivalents are listed in Table 1. The use of any benzodiazepine for treatment of AWS, regardless of the dose, was also compared between the treatment groups. As complica-
3 Mayo Clin Proc, July 2001, Vol 76 Alcohol Withdrawal in Inpatients 697 Table 1. Benzodiazepine Drugs Administered and Dosage Equivalents* Dosage equivalent Generic name Trade name (mg) Alprazolam Xanax Chlordiazepoxide PO Librium PO 25 Chlordiazepoxide IV Librium IV 12.5 Chlordiazepoxide IM Librium IM 12.5 Clonazepam Klonopin 0.25 Clorazepate Tranxene 15 Diazepam Valium 10 Flurazepam Dalmane 15 Halazepam Paxipam 40 Lorazepam PO Ativan PO 1 Lorazepam IV Ativan IV 0.5 Oxazepam Serax 30 Prazepam Centrax 10 Temazepam Restoril 15 Triazolam Halcion 0.25 *IM = intramuscular; IV = intravenous; PO = oral. tions and adverse outcomes of alcohol withdrawal were infrequent, we assessed differences between cohorts for only 2 end points: the occurrence of delirium tremens and the presence of any complication or adverse outcome. We defined delirium tremens as alcohol withdrawal accompanied by disorientation, agitation, and hallucinations as described in physician progress notes from patient histories. The associations between treatment group and benzodiazepine use and alcohol withdrawal complications and adverse outcomes were assessed with use of logistic regression models. Generalized estimating equations were used to fit these models since a number of patients were admitted for treatment of alcohol withdrawal multiple times. Generalized estimating equations, developed by Liang and Zeger, 9 were used to account for the correlation among repeated measurements taken from a single individual. The associations between treatment group and duration of treatment and benzodiazepine equivalents were assessed with mixed-models analyses of covariance (ANCOVA), again accounting for multiple admissions in the same patient. To meet the model-fitting assumptions required by the ANCOVA models, duration of treatment and total benzodiazepine equivalents were analyzed on the square root and natural logarithm scales, respectively. Because this study was retrospective in nature and the treatment groups were separated by time as well as by treatment, we attempted to account for differences between the treatment groups due to factors other than the treatment of interest. Factors that could have affected the differences between the treatment groups included Charlson comorbidity index, age, sex, previous alcohol withdrawal, previous episodes of delirium tremens, and previous alcohol withdrawal seizures. To control for possible differences due to these factors, we included them as covariates in all logistic regression and ANCOVA models. Significant differences in benzodiazepine use and complications of alcohol withdrawal were summarized with odds ratios and 95% confidence intervals. P values less than.05 were considered statistically significant. RESULTS Patient and Alcohol Withdrawal Treatment Characteristics Of 638 admissions found (536 patients), 216 met the inclusion criteria for this study (Table 2). The patient and alcohol withdrawal treatment characteristics are summarized by admission treatment group in Tables 3 and 4. There were 63 male admissions (75.0%) for AWS in the preimplementation cohort, and 95 (72.0%) were postim- Table 2. Exclusion Criteria by Admission Treatment Group* No. (%) excluded by treatment group Exclusion criteria (n=306) (n=332) ICU at initiation of treatment 81 (26.5) 63 (19.0) Not on primary medical service 73 (23.9) 60 (18.1) Pregnancy 0 (0.0) 0 (0.0) Age <18 y 0 (0.0) 0 (0.0) Multiple drug overdose 0 (0.0) 0 (0.0) Seizure prior to initiation of treatment 30 (9.8) 37 (11.1) Data unavailable 5 (1.6) 3 (0.9) Alcohol withdrawal not present 33 (10.8) 27 (8.1) CIWA not used for alcohol withdrawal NA 10 (3.0) Eligible 84 (27.5) 132 (39.8) *CIWA = Clinical Institute Withdrawal Assessment; ICU = intensive care unit; NA = not applicable.
4 698 Alcohol Withdrawal in Inpatients Mayo Clin Proc, July 2001, Vol 76 Table 3. Patient and Alcohol Withdrawal Characteristics by Admission Treatment Group Treatment group Characteristics (n=84) (n=132) No. (%) male 63 (75.0) 95 (72.0) No. (%) with previous alcohol withdrawal 52 (63.4) 78 (59.5) No. (%) with previous withdrawal seizures 12 (14.8) 19 (14.5) No. (%) with previous delirium tremens 19 (23.5) 15 (11.5) Age at admission (y) Mean (SD) 55.4 (15.5) 51.8 (13.7) Median (range) 54.9 ( ) 48.4 ( ) Charlson comorbidity index Mean (SD) 2.3 (2.5) 1.8 (2.0) Median (range) 1.5 (0-10) 1.0 (0-11) plementation admissions. The average (SD) age at admission was 55.4 (15.5) years vs 51.8 (13.7) years in the 2 cohorts. The median length of hospitalization was 3 days for both cohorts. Median Charlson comorbidity index was 1.5 and 1.0 for the preimplementation and postimplementation cohorts, respectively. The preimplementation cohort included 52 admissions (63.4%) of patients with documented previous episode(s) of AWS compared with 78 admissions (59.5%) of patients with previous AWS in the postimplementation cohort. Previous episodes of delirium tremens were documented in 19 preimplementation admissions (23.5%) and 15 postimplementation admissions (11.5%) (P=.02). Duration of Alcohol Withdrawal Treatment The average (SD) duration of alcohol withdrawal treatment with benzodiazepine was 55.5 (54.5) hours in the preimplementation admissions and 44.9 (49.6) hours in the postimplementation admissions. After adjustment for Charlson comorbidity index, age, sex, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens, there was no evidence of a difference in duration of treatment between the 2 cohorts (P=.16). Benzodiazepine Use Benzodiazepine was prescribed in 74 preimplementation admissions (88.1%) and 108 postimplementation admissions (81.8%). Chlordiazepoxide and lorazepam were the benzodiazepine formulations used most often in both cohorts, accounting for 97% of all doses administered in the preimplementation cohort and 99% of all doses administered in the postimplementation cohort. After adjustment for Charlson comorbidity index, age, sex, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens, there was no significant difference in benzodiazepine use between usual care and symptom-triggered therapy (P=.21). Other drugs for the treatment of AWS were used infrequently in both admission cohorts. β-blockers were used in 8 preimplementation admissions (9.9%) and in 26 postimplementation admissions (19.7%). α-adrenergic agonists were used in 2 preimplementation admissions (2.6%) and 5 postimplementation admissions (3.9%). One preimplementation admission (1.3%) and 4 postimplementation admissions (3.1%) received anticonvulsant medication. No admission in either cohort received intravenous alcohol. Total Benzodiazepine Equivalents Among patients who received any benzodiazepine, the average (SD) total benzodiazepine equivalents given during treatment for alcohol withdrawal symptoms was 20.1 mg (20.7 mg) and 20.1 mg (29.7 mg) in the 2 admission cohorts. After adjustment for Charlson comorbidity index, age, sex, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens, there was no evidence of a difference in total benzodiazepine equivalents between the 2 cohorts (P=.38). In addition, after adjustment for the covariates listed previously, there was no difference in total dose of benzodiazepine on day 1 of therapy (P=.27). Complications Delirium tremens was observed in 17 preimplementation admissions (20.5%) and 9 postimplementation admissions (6.9%). The difference in observed delirium tremens between the 2 cohorts after adjustment for Charlson comorbidity index, age, sex, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens was significant (P=.04) (Table 5). Symptom-triggered
5 Mayo Clin Proc, July 2001, Vol 76 Alcohol Withdrawal in Inpatients 699 Table 4. Alcohol Withdrawal Outcomes by Admission Treatment Group Treatment group Outcome (n=84) (n=132) No. (%) with benzodiazepine prescribed 74 (88.1) 108 (81.8) Duration of treatment (h) No.* Mean (SD) 55.5 (54.5) 44.9 (49.6) Median (range) 38.9 ( ) 31.8 ( ) Total benzodiazepine equivalents No.* Mean (SD) 20.1 (20.7) 20.1 (29.7) Median (range) 10.8 ( ) 9.0 ( ) *Number of admissions for which this outcome was calculated. therapy seemed to be most effective in preventing delirium tremens in patients without a prior history of delirium tremens. Of those preimplementation admissions without a history of delirium tremens, 16% had an episode during the admission. Only 2.6% of postimplementation admissions with no history of delirium tremens had a current episode. However, of patients with a history of previous delirium tremens, 36.8% and 40.0% in the respective cohorts had an episode during the current admission. This interaction between cohorts and prior history of delirium tremens was statistically significant (P=.03). A complication or adverse outcome of any kind was observed in 27 preimplementation admissions (32.5%) and 23 postimplementation admissions (17.6%). The preimplementation admissions included 12 (14.8%) with a history of previous AWS seizures compared with 19 (14.5%) in the postimplementation cohort. The difference in complication rates between the 2 cohorts was not statistically significant after adjustment for Charlson comorbidity index, age, sex, previous AWS, previous alcohol withdrawal seizures, and previous delirium tremens (P=.053), although a trend for fewer complications was observed in the postimplementation cohort. DISCUSSION Benzodiazepines provide effective treatment for alcohol withdrawal symptoms and also decrease the risk of seizures and delirium tremens in placebo-controlled studies. 1 Symptom-triggered therapy produces equivalent reduction in symptoms compared with fixed-schedule benzodiazepine treatment in chemical dependency unit patients. 6 However, little data exist to demonstrate the superiority of symptomtriggered therapy for treating the symptoms and complications of AWS over traditional strategies of benzodiazepine use in medical inpatients. Sullivan et al 10 reported less benzodiazepine use in a general hospital setting for patients admitted with alcohol dependency and managed with symptom-triggered therapy compared with usual care, but they reported no difference in withdrawal complications. Our study suggests that, for patients admitted to general medical services who experience alcohol withdrawal, symptom-triggered treatment is associated with a reduced risk of delirium tremens compared with usual care. This reduction in risk was observed as a significantly lower rate of delirium tremens, particularly in patients who had no prior documented episode of delirium tremens. The incidence of delirium tremens in medical inpatients has been reported infrequently and was unknown to us in our care setting. Gerke et al 11 found 12.8% of patients admitted to a general hospital with alcohol-related diseases had delirium tremens. Ferguson et al 12 found an incidence of 24% in an indigent population treated for AWS or detoxification. There was independent correlation between delirium tremens and acute medical illness and an extended period since last consumption of alcohol. Thus, it appears that social factors and acute medical illness can influence the observed rate of delirium tremens. 13 The mean Charlson comorbidity index score for both the pre- and postimplementation cohorts in our study reflected underlying medical illness. Thus, our alcohol-dependent population was a cohort with higher risk for AWS and more complicated AWS than patients with alcohol dependence without medical illness. We excluded patients with seizures at presentation or the need to be admitted to the intensive care unit. Inclusion of these patients would have resulted in even higher acuity of medical illness with higher rates of withdrawal complications, not typical of general medical patients. Other investigators have demonstrated that a symptomtriggered approach results in decreased duration of benzodiazepine use and increased rates of observation without pharmacologic therapy compared with a fixed-schedule
6 700 Alcohol Withdrawal in Inpatients Mayo Clin Proc, July 2001, Vol 76 Table 5. Alcohol Withdrawal Complications and Adverse Outcomes by Admission Treatment Group* No. (%) by treatment group Complication/adverse outcome (n=84) (n=132) Delirium tremens 17 (20.5) 9 (6.9) Seizures 0 (0.0) 0 (0.0) Readmission for withdrawal symptoms 0 (0.0) 3 (2.3) Left hospital against medical advice 5 (6.0) 4 (3.1) Transfer to ICU for withdrawal 0 (0.0) 0 (0.0) Transfer to ICU for other reason 5 (6.0) 7 (5.3) Death from withdrawal 0 (0.0) 0 (0.0) Death from other causes 2 (2.4) 0 (0.0) Any complication/adverse outcome 27 (32.5) 23 (17.6) *ICU = intensive care unit. strategy. 6 We did not observe this association in part because coexistent medical illness can amplify withdrawal symptoms, leading to longer benzodiazepine use. Our study has several potential weaknesses inherent in any retrospective analysis. We identified patients in this study through both discharge diagnoses and pharmacy records verifying patient receipt of benzodiazepines and thiamine. It is possible that these criteria affected the cohort composition, especially for patients who were never treated pharmacologically. Duration and use of pharmacologic treatment may have been biased against the postimplementation cohort; however, a larger, though not significant proportion of this group did not receive benzodiazepines. Our intervention may have led to more uniform use of thiamine or more frequent coding of AWS as a discharge diagnosis. Although we attempted to account for variables that would affect outcomes, other possible confounders include the exact nature of comorbid medical illness. Importantly, the ascertainment of the occurrence of delirium tremens was less reliable than in a prospective study. Although we used standard criteria to define delirium tremens, they were applied by a nonblinded abstractor who was dependent on documentation in the medical record. However, few patients in either treatment group received other interventions that might have affected the outcome of treatment for AWS. Our observation of a decrease in the risk of delirium tremens, particularly for patients without a previously documented episode, could be explained by 2 factors. First, our intervention provided for close observation for withdrawal symptoms by registered nurses using a validated scoring tool; while not altogether different from usual care, these observations were more consistent and more uniform. This level of observation could have resulted in more frequent and more adequate dosing of medication. Also, front loading benzodiazepine therapy could have a positive effect on the outcome of a withdrawal episode. Although the dose of benzodiazepine was not different between cohorts in the first 24 hours of treatment, there may have been higher dosage in the first 12 hours of the postimplementation cohort as prescribed by our protocol. We were not able to ascertain dosing for intervals of less than 24 hours. Altogether, our intervention attempted to provide optimal matching of pharmacologic therapy to symptoms. The lack of a decrease in the occurrence of delirium tremens for patients with a documented prior episode supports the hypothesis that this syndrome has strong determinants in the central nervous system adaptation to heavy, long-term alcohol exposure. The actual amount of exposure necessary to result in delirium tremens is not definable. However, consistent with the known reduction in delirium tremens with benzodiazepine treatment, it is highly plausible that a strategy that optimally provides benzodiazepine replacement may have considerable impact on patients whose risk for delirium tremens is more modest. In summary, we have shown in this retrospective analysis that medical inpatients experiencing AWS can be treated effectively with use of symptom-triggered therapy. Although benzodiazepine use and duration of treatment did not differ in our treatment groups, the most serious and lifethreatening complication of AWS, delirium tremens, occurred significantly less frequently with symptom-triggered therapy, especially when it was prescribed for patients with no prior history of delirium tremens. Further studies with larger numbers of patients will help clarify the difference we observed for the incidence of delirium tremens. The authors gratefully acknowledge Robert M. Morse, MD, for review of the manuscript, Christine M. Lohse, BS, for statistical expertise, and Luann Koenig, RN, for assistance in abstraction of data.
7 Mayo Clin Proc, July 2001, Vol 76 Alcohol Withdrawal in Inpatients 701 REFERENCES 1. Mayo-Smith MF, American Society of Addiction Medicine Working Group on Pharmacological Management of Alcohol Withdrawal. Pharmacological management of alcohol withdrawal: a meta-analysis and evidence-based practice guideline. JAMA. 1997;278: Saitz R, Friedman LS, Mayo-Smith MF. Alcohol withdrawal: a nationwide survey of inpatient treatment practices. J Gen Intern Med. 1995;10: Sereny G, Kalant H. Comparative clinical evaluation of chlordiazepoxide and promazine in treatment of alcohol-withdrawal syndrome. BMJ. 1965;1: Sellers EM, Naranjo CA, Harrison M, Devenyi P, Roach C, Sykora K. Diazepam loading: simplified treatment of alcohol withdrawal. Clin Pharmacol Ther. 1983;34: Kaim SC, Klett CJ, Rothfeld B. Treatment of the acute alcohol withdrawal state: a comparison of four drugs. Am J Psychiatry. 1969;125: Saitz R, Mayo-Smith MF, Roberts MS, Redmond HA, Bernard DR, Calkins DR. Individualized treatment for alcohol withdrawal: a randomized double blind controlled trial. JAMA. 1994;272: Sullivan JT, Sykora K, Schneiderman J, Naranjo CA, Sellers EM. Assessment of alcohol withdrawal: the revised Clinical Institute Withdrawal Assessment for Alcohol scale (CIWA-Ar). Br J Addict. 1989;84: Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73: Sullivan JT, Swift RM, Lewis DC. Benzodiazepine requirements during alcohol withdrawal syndrome: clinical implications of using a standardized withdrawal scale. J Clin Psychopharmacol. 1991; 11: Gerke P, Hapke U, Rumpf HJ, John U. Alcohol-related diseases in general hospital patients. Alcohol Alcohol. 1997;32: Ferguson JA, Suelzer CJ, Eckert GJ, Zhou XH, Dittus RS. Risk factors for delirium tremens development. J Gen Intern Med. 1996; 11: Schuckit MA, Tipp JE, Reich T, Hesselbrock VM, Bucholz KK. The histories of withdrawal convulsions and delirium tremens in 1648 alcohol dependent subjects. Addiction. 1995;90: