Iatrogenic Diseases as a Reason for Admission to the Intensive Care Unit:  Incidence, Causes, and Consequences FREE

THERE HAVE been numerous North American and European studies on iatrogenic diseases (IDs) during the past 3 decades. Although the incidence of community-acquired ID is not precisely known, it seems to account for 2% to 10% of outpatient consultations^1- 2 and 3% to 7% of hospital admissions,^3- 10 most of them (95%) after drug exposures. These patients spend 8 to 10 days in the hospital and have a fatality rate of 2% to 6% and a rate of preventable events of 50%.^1,4- 6,10 The elderly are most at risk.^2,6,10- 11 Hospital-acquired ID occurs in 2% to 36% of hospitalized patients^6,12- 19 and can be caused by drug-induced illness or events after medical and surgical procedures (35%-75% of cases). Hospital-acquired ID increases the length of stay and has a fatality rate of 3.7% to 14.0% and a permanent disability rate of 6.5%.^10,14 The rate of preventable events is 20% to 50%.^10,14- 16 Once again, the elderly are more at risk,^{6,10- 11,13,17- 18} but the severity of the underlying diseases,^5,20- 22 the number of prescribed drugs,^13,23 and the pattern and location of care^14- 15,24 are also factors. The risk of ID is high in neurologic, thoracic, vascular, and cardiac surgery units; intensive care units (ICUs); emergency departments; and interventional radiology because the association of severe coexisting diseases and the complexity of treatment are likely to lead to iatrogenic events.

Severity is closely linked to the nature of the adverse event and the underlying medical condition. Iatrogenic complications are life-threatening in 10% to 26% of cases.^{13,15,17- 18,24} Trunet et al²⁵ found that ID accounted for 12.6% of admissions in the ICU, with a fatality rate of 20.0%, in an adult medicosurgical ICU affiliated with a French tertiary care university hospital.

Most of the studies published during the past 25 years have shown the negative effects of ID on population health, the cost of medical care,^26- 27 and the rate of malpractice claims. Health care providers have reacted by developing quality-of-care initiatives,^28- 30 including better identification of ID,^31- 34 determination of risk factors,^{14- 16,24,35- 38} and studies on the epidemiology of malpractice claims.^39- 43 Nevertheless, medicine has changed considerably during the past few decades, with general aging of the population and the development of increasingly complex or advanced procedures, such as invasive cardiology and radiology and laparoscopic surgery. We have attempted to assess the impact of these changes on ID in a retrospective study. We compared the rate of ID-related admissions with the results reported by Trunet et al²⁵ 15 years ago, and analyzed the causes, risk factors, and consequences of these diseases.

This study was conducted in the 15-bed medicosurgical ICU (MSICU) affiliated with the 500-bed General Hospital at Compiègne, France. Cardiac surgery, neurosurgery, and organ transplantation are not performed in this institution. In addition to the MSICU, a 6-bed cardiac care unit takes care of patients with cardiovascular emergencies in the absence of other severe organ failure. Admission to the MSICU of any patient with at least 1 life-threatening organ failure or metabolic disorder, or requiring emergency dialysis or mechanical ventilation, is authorized by 1 of the permanent or on-duty MSICU senior physicians after discussion with the physician (emergency department, operating room, or wards) caring for the patient. All the patients admitted to the MSICU between January 1, 1994, and December 31, 1994, were retrospectively screened for ID as the purpose of admission to the unit. Screening was performed by house staff composed of 3 board-certified anesthesiologists (B.D., E.L.M., E.B.), 1 cardiology and internal medicine board-certified intensivist (Y.D.), and 1 resident (B.F.). Each patient was identified as having ID or not by consensus, and the causes, relationship between prescription or procedure and ID, preventability, and severity of the iatrogenic events were determined. If consensus was not obtained, the patient was not included. Each assessment was made according to the following criteria and definitions.

Iatrogenic disease was defined as a disease induced by a drug prescribed by a physician; or after a medical or surgical procedure, excluding intentional overdose, nonmedical intervention; or unauthorized prescription, and environmental events (falls, equipment defect).

Adverse event was defined as an unintended and noxious event caused by medical management carried out according to the best of medical science.

Preventable event was defined as an event that should not occur if management is the best that medical science can provide.

Nosocomial infection was defined as a localized or systemic infection, occurring at least 48 hours after hospital admission, that was not present or incubating at the time of admission.⁴⁴

Iatrogenic infection was defined as an infection after medical or surgical management, whether or not the patient was hospitalized.

We used the progressive criteria of Karch and Lasagna.⁴⁵ The relationship between a drug and an adverse reaction was divided into the following 5 levels: definite: a reaction that follows a reasonable temporal sequence from administration of the drug, or in which the drug level has been established in body fluids or tissues, that follows a known response pattern to the suspected drug, and that is confirmed by dechallenge and rechallenge; probable: a reaction that follows a reasonable temporal sequence from administration, that follows a known response pattern to the suspected drug, that is confirmed by dechallenge, and that could not be reasonably explained by the the patient's condition; possible: a reaction that follows a reasonable temporal sequence from administration and that follows a known response pattern to the suspected drug, but that could have been produced by the patient's clinical condition or other therapy given to the patient; conditional: a reaction that follows a reasonable temporal sequence from administration and that does not follow a known response pattern to the suspected drug, but that could not be reasonably explained by the known characteristics of the patient's clinical condition; and doubtful: any reaction that does not meet the above criteria.

As there is no standard definition, we used the same criteria as Trunet et al.²⁵ The relationship was good if the following 4 criteria were all satisfied; otherwise the relationship was conditional: the complication is known and recorded in the medical literature; it is not reasonably explained by the patient's underlying diseases; there is a reasonable temporal sequence from the procedure to the complication; and a relationship can be established from anatomical criteria (such as colonic perforation linked to endoscopy), microbiologic criteria (such as urinary tract infection after urinary catheterization), or chemical criteria (such as radiocontrast and renal failure).

The severity of the ID was classified as fatal, life threatening, or moderate. A fatal event is a complication principally responsible for death. Life-threatening ID requires intensive care (mechanical ventilation, vasopressors, hemodialysis, cardiac catheterization or pacing, tube thoracostomy, surgery, etc). Complication was judged to be moderate if only routine management and monitoring were needed.

The following data were recorded for each patient admitted to the ICU during the study period: age, sex, number of prescribed drugs before admission to the ICU, origin of patient (home or hospital), length of stay, and outcome in the ICU. The Simplified Acute Physiology Score⁴⁶ was calculated within 24 hours after admission to the ICU as an index of disease severity. The McCabe score⁴⁷ (3 classes: fatal during hospitalization, ultimately fatal within 5 years, and not fatal) was used as an index of the severity of the underlying medical condition. The Omega score was calculated at the end of the patient's stay as an index of total workload (Figure 1). The Omega score system is an ICU-specific activity scoring system used in the French ICUs since 1990, validated by the French Health Department and the Société de Réanimation de Langue Française.⁴⁸ It is highly correlated with the Nursing Research Project 1987 scoring system used in Canada. The Omega score is calculated by summing up 47 therapeutic interventions collected during each patient's ICU stay and scored from 1 to 10. For instance, tracheostomy is scored as 6 points per hospitalization; hemodialysis, 10 points per course; and mechanical ventilation, 10 points per day. We also recorded the nature and relationship between disease and procedure, and the severity and the preventability of disease for each patient with ID. The cost of care in the ICU was calculated by means of the global "day's fare" allocated to our ICU by the French care system, which was Fr 8122 per day for 1994 (US $1500 per day).

The characteristics of the patients are expressed as means±SDs or as the number of patients or events. Comparisons between patients were made by Student t test for continuous variables and the χ² test with Yates correction when indicated for categorical variables. A P value of less than .05 was considered statistically significant. Data were stored on Microsoft EXCEL (Version 4.0; Microsoft Inc, Redmond, Wash), and all probability calculations were performed with SAS Macro software (SAS Institute Inc, Cary, NC).

A total of 24,555 patients were hospitalized at the General Hospital in Compiegne during 1994. Of these, 16,746 went to adult medical and pediatric wards, 7186 to surgical and obstetric wards, and 623 to the adult medicosurgical ICU. Among these 623 patients, 68 were admitted to the ICU because of 68 IDs, accounting for 10.9% of the ICU admissions. Twenty-seven (40%) were referred from home and 41 (60%) from hospital wards. The mean age was 69.4±17.2 years, the sex ratio (male-female) was 0.66, and the mean Simplified Acute Physiology Score was 13.4±6.7. The ID cases included 41 patients with drug-related ID (Table 1), 12 patients whose ID was linked to medical procedures (Table 2), and 15 whose ID was linked to surgical procedures (Table 3). Iatrogenic disease was life-threatening for 14 patients (21%) and fatal for 9 patients (13%). The rate of preventable ID was 51%.

The 41 drug-induced ID cases accounted for 60% of ID admissions to the ICU (24 from home, 27 from the hospital) and included 11 adverse events and 30 preventable events. The symptoms were severe in 11 instances and resulted in death in 6 cases. The relationship between drug and ID was definite in 14, probable in 14, and possible in 13. Fifty-five drugs were involved. Morbid combinations (ie, angiotensin-converting enzyme inhibitors±nonsteroidal anti-inflammatory drugs±diuretics) were encountered in several instances, but a single drug (antibiotic, vitamin K antagonist, etc) was responsible for ID in 22 cases (Table 4).

Twelve patients were admitted for ID after a medical procedure (3 performed out of hospital). The relationship was good in all cases, and ID was preventable in 3 of the 12 cases. Symptoms were severe in 7 instances, resulting in death in 2 cases. Six patients had iatrogenic infections, among which 3 were nosocomial infections and 3 were community-acquired infections (tetanus, bacteremia after lower-limb venous sclerotherapy and urinary catheter).

Fifteen patients had surgical complications, all of which occurred in the operating theater, in the recovery room, or on surgical wards. The relationship was good in 14 of the 15 cases, and the complication was preventable in 2 instances. Symptoms were severe in 5 cases, leading to death in 1 patient.

Overall, these 68 admissions resulted in 472 days of hospitalization in the ICU, with a mean length of stay of 6.9±9.3 days (range, 1-52 days). The workload required was 3612 Omega points, with a mean of 53 points per patient and a density rate of 7.7 points per day of hospitalization. The fatality rate in the ICU was 13% (9/68 patients). Cost of medical care in the ICU for these 68 patients was estimated at US $688,470.

About half (35/68) of these IDs were considered to be preventable. The 30 preventable events after drug-induced ID were caused by the use of inappropriate drugs in 6 cases, an error in dose in 8 cases, inadequate follow-up of therapy in 14 cases, and failure to use prophylactic treatment in 2 cases. Three events were linked to medical procedures (radiocontrast infusion in 2 and hemorrhoid sclerotherapy in 1). Two were caused by surgical procedures (suprapubic cystotomy tube and epistaxis tamponade). These 35 preventable IDs resulted in 189 days of hospitalization in the ICU and 5 deaths (14%). Cost of medical care in the ICU was evaluated at US $275,680, with a workload of 1434 Omega points (41 points per patient, 7.6 points per day).

The 623 patients admitted to the ICU during the study period were divided into group 1 (n=555), patients without IDs; group 2 (n=41), patients with drug-induced IDs; and group 3 (n=27), patients with IDs after medical or surgical procedures (Table 5). Groups 2 and 3 were next compared individually with group 1. Group 2 patients were older and had an inverted sex ratio (29 women and 12 men), a higher Simplified Acute Physiology Score, and more prescribed drugs on admission. Group 3 patients were also older and stayed longer in the ICU. Nevertheless, groups 2 and 3 did not differ from group 1 with respect to workload in the ICU (Omega score) or mortality. Likewise, patients who experienced adverse events were compared with those who experienced preventable events (Table 6). The only factor increasing the risk of preventable events was age. The location of drug prescription (home or hospital) was not a risk factor for drug-induced ID.

This study, conducted 15 years after the one by Trunet et al,²⁵ clearly confirms the persistence and the paramount importance of ID as a cause of admission to the ICU. Trunet et al found that 12.6% of the ICU admissions were linked to ID. The rate was 10.9% in the present study, with the use of the same criteria and definitions. There was also a high rate of preventable events, accounting for 51% of the ID in our study, while Trunet et al found 41%. The stability of these rates over this long time is somewhat surprising, as a decrease might have occurred because of better recognition, care, and prophylaxis for ID or an increase because of more powerful and/or invasive treatment for more sick and aged patients. The 2 trends may well have canceled each other out.

We believe that this exhaustive review, although retrospective, of all ICU charts and medical reports from the 623 patients admitted during 1994, done early in the following year by the permanent clinical team taking care of these patients, provides a sensitive investigation. However, this advantage is probably outweighed by the method of identification of ID, requiring unanimity of judgment for a definitive inclusion. Although this procedure provided reliability of assessments, we believe that it led to an underestimation of the number of included IDs.

The drugs implicated in ID remained standard. Cardiovascular drugs accounted for 31%, anti-inflammatory and analgesic drugs for 20%, and antibiotics for 11% of cases of drug-induced ID. This has not changed in 20 years,^4- 5 but there have been striking changes in each class of drugs.

Angiotensin-converting enzyme inhibitors (alone or associated with diuretics) are the leading class of cardiovascular drugs involved in ID, before diuretics and oral anticoagulants, whereas digitalis compounds and older antihypertensive drugs have almost disappeared. Few IDs related to antiarrhythmic or thrombolytic drugs were encountered in this study, because of the presence of a 6-bed cardiac care unit in the same hospital. Nonsteroidal anti-inflammatory drugs emerged as the leading anti-inflammatory and analgesic drugs causing iatrogenic events, and there were no complications linked to corticosteroids or aspirin. These changes probably reflect changes in prescription habits.

Trunet et al²⁵ found 2 incompatible blood transfusions among the 23 drug-induced IDs, whereas we observed no complication linked to the transfusion of blood products. The drastic vigilance guidelines regarding blood products enforced in France in 1992 after the "contaminated blood products affair" have reduced the frequency of gross compatibility accidents.⁴⁹ We also saw no severe ID caused by antidiabetic drugs, perhaps as a result of improved education of diabetic patients by the endocrinological staff of our hospital. Finally, only 2 of the 41 patients with drug-induced ID had allergic reactions severe enough to require ICU admission. This recruitment bias explains the differences of distribution for drug-induced ID, when compared with out-of-hospital or ward studies.^{6,10,13,15,17}

The 12 IDs caused by medical procedures included 6 severe infections, 3 of them hospital acquired (2 cases of bacteremia after peripheral venous catheterization and 1 after urinary tract catheterization). We could not assess the preventability of these catheter-related infections because we had no information about their real need and the adherence to aseptic guidelines for their insertion and care.

The preventability of an iatrogenic accident after surgery is often difficult to determine because the surgeon is both judge and judged. This probably lowers the ability to distinguish between an adverse event and a preventable event. The 2 surgical iatrogenic events defined as preventable included (1) a laborious attempt at suprapubic cystotomy drainage, without previous echographic verification of the bladder repletion, leading to a bladder laceration, and (2) a mispositioned nasopharyngeal balloon tamponade (inflated through the larynx) for active epistaxis, which was followed by acute respiratory distress. These 2 procedures were performed by residents, with a delayed call to senior staff because of the disaster.

The overall fatality rate of patients with ID was 13% (9/68 patients), not significantly different (χ²=0.55; P>.50) from the 17% fatality rate of the patients admitted to the ICU for other reasons (93/555). The fatality rate in the study by Trunet et al²⁵ was 19.5% (not significantly different from our study [χ²=0.76; P>.50]).

The risk factors identified herein are the same as in earlier studies ^{6,10- 11,13,17- 18}: old age (probably mainly women in our country) and the number of drugs prescribed before admission.^13- 14,23 Drug-induced iatrogenic events, and particularly the preventable ones, are as likely to occur at home as in the hospital.

We distinguished preventable iatrogenic events from adverse events because the implications for responsibility of physicians are obviously quite different. An adverse event is an unexpected, unavoidable deleterious event after a medical prescription or procedure. That is the price we pay for using powerful drugs and sophisticated procedures. These modern medical tools may result in an overall increase in life expectancy, but the individual benefit-risk ratio decreases in older patients, patients with organ insufficiency, and patients receiving multiple medications.^{11,13,20- 23} Steel et al¹³ and Ponge et al⁵⁰ have shown that the multiple drugs prescribed for patients hospitalized for drug-induced ID could be reduced by 25% to 40% without further damage.

A preventable event is an event that could have been avoided if the medical act or prescription had respected the state of the art of medical science. The 51% rate of preventable events (35/68 patients) found in our study agrees with the results of others.^{10,14- 16,25,32,51} This high rate results in part from biased recruitment to the ICU, as the severity of iatrogenic disease seems to be correlated with errors.¹⁴ The immediate causes of preventable ID recorded are unremarkable^15,18,25: dosage error, contraindication or drug interaction unawareness, prophylaxis or monitoring omission, and technical error. However, the in-depth underlying causes of human fallibility and malpractice leading to these mishaps are worth examining. There are 3 types of error: error resulting from ignorance, negligence, or misjudgment.

Errors caused by ignorance (2 of 35 errors in this study) might be illustrated by a physician who does not recognize an unusual disease or does not perform exceptional care, although he or she took attentive care of the patient according to the best of his or her own knowledge. Such errors are probably difficult to avoid as the complexity and field of medical knowledge increase.

Errors caused by negligence (22 of 35 errors in this study) include acts that are knowingly below acceptable standards, as are violations of aseptic guidelines or omission of allergy checking. This lack of rigor probably is prompted by multiple factors, such as exhaustion, business factors, and loss of motivation.

Couch et al¹² originally identified 5 causes of medical misjudgment in surgery, easily transposable to any medical activity: (1) misplaced optimism, (2) unwarranted urgency, (3) an urge for perfection, (4) fashionable therapy, and (5) insufficient restraint and deliberation.

The cost of ID for these 68 patients in human (472 days in the ICU, 9 deaths) and financial (US $688,470) terms is probably underestimated. The financial costs were calculated by means of a daily set price calculated and revised each year by the hospital financial services. The morbidity, sequelae, and later morbidity attributable to ID after the survivors left the ICU were not followed up. If the present study group (68 patients per year per 200,000 inhabitants) is considered to be representative, extrapolation leads to 45 deaths, 2360 days of hospitalization in the ICU, and US $3.44 million per 1 million inhabitants per year. These amazing figures do not include IDs occurring in other high-risk units not present in our hospital and cases treated outside the ICU, at home, or unrecognized. Indirect iatrogenic events, such as crashes and occupational or domestic accidents experienced by patients who are prescribed psychotropic drugs, are also not recorded as IDs. The type of severe iatrogenic events we investigated may well be only the tip of the ID iceberg.

We confirmed that ID is still a major cause of hospitalization in the ICU, accounting for 11% of admissions, half of which are preventable. The morbidity, mortality (13% in our study), and cost of ID are certainly biased by the design of this study (patients with ID severe enough to require treatment in the ICU). Nevertheless, the global impact of ID on the health and resources of society is probably underestimated. The challenge, in this era of high-technology medicine, is to better understand the in-depth underlying causes of errors, to reduce the incidence of preventable iatrogenic events. This could provide a better quality of care at a lower cost.

Accepted for publication April 30, 1998.

Reprints: Bruno Darchy, MD, Service de Réanimation, Centre Hospitalier—Zac de Mercieres, 8, avenue Henri Adnot, 60200 Compiegne, France.