Predictions of hospital mortality rates: a comparison of data sources.

BACKGROUND: Comparing hospital mortality rates requires accurate adjustment for patients' intrinsic differences. Commercial severity systems require either administrative data that omit vital clinical facts about patients' conditions at hospital admission or costly, time-consuming abstraction of medical records. The validity of supplementing administrative data with laboratory data has not been assessed. OBJECTIVE: To compare risk-adjusted mortality predictions using administrative data alone; administrative data plus laboratory values; and the combination of administrative, laboratory, and clinical data. DESIGN: Retrospective cohort study. SETTING: 30 acute care hospitals. PATIENTS: 46,769 patients hospitalized with acute myocardial infarction, cerebrovascular accident, congestive heart failure, or pneumonia. MEASUREMENTS: Each patient's probability of dying was estimated by using administrative data only (unrestricted administrative models), administrative data restricted to secondary diagnoses that are unlikely to be hospital-acquired complications (restricted administrative models), restricted administrative data plus laboratory data (laboratory models), and restricted administrative data plus laboratory and abstracted clinical data (clinical models). RESULTS: The unrestricted administrative models predicted death better than the restricted administrative models (average areas under the receiver-operating characteristic [ROC] curves, 0.87 and 0.75, respectively) and as well as the laboratory models and the clinical models (average areas under the ROC curves, 0.86 and 0.87, respectively). The good mortality predictions obtained by using the unrestricted administrative models result from inclusion of hospital-acquired complications that commonly precede death. The laboratory models ranked 93% of patients and 95% of hospitals in a manner similar to the clinical models; in comparison, rankings provided by the laboratory models were similar to those provided for 75% of patients and 69% of hospitals by the unrestricted administrative models and for 72% of patients and 77% of hospitals by the restricted administrative models. CONCLUSIONS: Adding laboratory data (often available electronically) to restricted administrative data sets can provide accurate predictions of inpatient death from acute myocardial infarction, cerebrovascular accident, congestive heart failure, or pneumonia. This alternative avoids the cost of data abstraction and the serious errors associated with using administrative data alone.

Course of antidepressant treatment with tricyclic versus selective serotonin reuptake inhibitor agents: a comparison in managed care and fee-for-service environments.

We compared course of treatment with tricyclic antidepressant drugs (TCADs) and selective serotonin reuptake inhibitors (SSRIs) to assess interactive effects of antidepressant type with payer type and patient characteristics. A nationwide sampling of adults (n = 4,252) from approximately equal numbers of health maintenance organization (HMO) and indemnity enrollees were prescribed no antidepressants for 9 months, and thereafter prescribed a TCAD or SSRI. Using a retrospective analysis of prescription claims, these cohorts of TCAD and SSRI utilizers were followed for 13 to 16 months after their initial antidepressant prescription. Outcome measures included (1) termination of antidepressant treatment before 1 month; and (2) failure to receive at least one therapeutic dose during treatment lasting 3 months or more. Rates of premature termination and subtherapeutic dosing were significantly higher for TCAD-treated than SSRI-treated patients, and for HMO than indemnity enrollees. The interaction of HMO enrollment and TCAD use was associated with particularly high rates. Excluding patients terminating in the first month, the proportions of TCAD and SSRI utilizers remaining in treatment over time were not significantly different. We conclude that SSRIs may provide advantages in treatment adherence and therapeutic dosing, particularly in environments with limited prescriber time. The first month of treatment may be especially critical in determining compliance.

Thermolabile methylenetetrahydrofolate reductase: an inherited risk factor for coronary artery disease.

Severe methylenetetrahydrofolate reductase (MTHFR) deficiency with less than 2% of normal enzyme activity is characterized by neurological abnormalities, atherosclerotic changes, and thromboembolism. We have discovered a "new" variant of MTHFR deficiency which is characterized by the absence of neurological abnormalities, an enzyme activity of about 50% of the normal value, and distinctive thermolability under specific conditions of heat inactivation. In this study, lymphocyte MTHFR specific activities in the thermolabile variant and control groups were 5.58 +/- 0.91 and 10.33 +/- 2.89 nmol formaldehyde formed/mg protein/h, respectively. The difference was significant (P less than .01). However, there was overlap among the individual values from the two groups. On the other hand, residual MTHFR activity after heat inactivation was 11.2 +/- 1.43% in the thermolabile variant and 36.3 +/- 5.18% in the controls. There was no overlap. Enzyme studies in 10 subjects with thermolabile MTHFR and their family members support the hypothesis that thermolabile MTHFR is inherited as an autosomal recessive trait. To elucidate the association of thermolabile MTHFR with the development of coronary artery disease, we determined the thermostability of lymphocyte MTHFR in 212 patients with proven coronary artery disease and in 202 controls without clinical evidence of atherosclerotic vascular disease. Thermolabile MTHFR was found in 36 (17.0%) cardiac patients and 10 (5.0%) controls. The difference in incidence between the two groups was statistically significant (P less than .01). The average age at onset of clinical coronary artery disease in 36 patients with thermolabile MTHFR was 57.3 +/- 7.6 years (35-72 years). The mean total plasma homocysteine concentration in patients with thermolabile MTHFR was 13.19 +/- 5.32 nmol/ml and was significantly different from the normal mean of 8.50 +/- 2.80 nmol/ml (P less than .05). There was no association between thermolabile MTHFR and other major risk factors. We conclude that thermolabile MTHFR is a variant(s) of MTHFR deficiency which is inherited as an autosomal recessive trait. In addition, it is positively associated with the development of coronary artery disease. Determination of in vitro thermostability of lymphocyte MTHFR is a reliable method for identifying subjects with this abnormality.

Acute renal failure associated with cardiac operations. A case-control study.

We performed a case-control study to identify risk factors for the development of acute renal failure after cardiac operations. Forty-two cases of acute renal failure were identified in a total of 572 patients who underwent cardiac operations. They were matched with a control population of patients having cardiac operations without acute renal failure. Discriminant analysis performed with preoperative variables revealed preoperative serum creatinine values, concurrent valve and bypass surgery, and age to be significant variables for identifying patients at risk for acute renal failure. The use of these three variables in a discriminant model correctly classified 77% of patients. The addition of intraoperative variables did not significantly improve the ability of the model to correctly classify patients. Acute renal failure was associated with a significant increase in the number of postoperative complications, mortality, and length of hospitalization and intensive care unit stay.

Uterine leiomyomas in the infertile patient: preoperative localization with MR imaging versus US and hysterosalpingography.

Eleven women with a history of infertility and uterine leiomyomas underwent magnetic resonance (MR) imaging of the pelvis prior to myomectomy. Nine also underwent preoperative pelvic ultrasonography (US), and ten underwent hysterosalpingography. All studies were interpreted prospectively by independent observers. With each imaging modality, the location (one of 11 anatomic segments), size, and appearance of detected uterine leiomyomas were determined and compared with surgical and histologic findings. Among the nine patients who underwent both MR and US, the sensitivity (85%) and accuracy (94%) of MR imaging for abnormal segments was significantly better than that of US (sensitivity = 69%, P = .015; accuracy = 87%, P = .043). For the ten patients who underwent both MR and hysterosalpingography, the sensitivity (91%) and accuracy (96%) of MR imaging was better than that of hysterosalpingography (sensitivity = 18%, P = .0005; accuracy = 72%, P = .0005). The specificities of the three modalities did not significantly differ (100%, 97%, and 98% for MR, US, and hysterosalpingography, respectively). These data suggest that MR imaging is superior to US or hysterosalpingography for preoperatively locating uterine leiomyomas.

Multiple cerebral metastases: detectability with Gd-DTPA-enhanced MR imaging.

Sixteen patients with suspected cerebral metastases were studied with magnetic resonance (MR) imaging before and after the intravenous administration of 0.1 mmol/kg of gadolinium diethylenetriaminepenta-acetic acid. The images were interpreted blindly by two neuroradiologists; all clinical, radiologic (computed tomographic and MR imaging), and pathologic data were reviewed to arrive at a final "best diagnosis," which was then compared with the prior blinded interpretations. Of seven patients found to have multiple metastases, six (86%) had at least one tumor nodule depicted by postinfusion MR imaging that was missed by one or both observers on review of preinfusion images alone. Lesions missed on preinfusion studies were usually small nodules hidden by or not detected next to regions of high-signal edema thought to be related to the adjacent tumor nodule. The authors believe that contrast enhancement improves detection of metastatic foci with MR imaging and that the findings indicate broader implications for the detection of multiple lesions from other causes.

Homocysteinemia due to folate deficiency.

We examined the relationship between serum folate and total homocyst(e)ine levels by determining protein-bound homocyst(e)ine in stored serum from 19 subjects with subnormal serum folate (less than 2 ng/mL), 137 subjects with low normal serum folate (between 2.0 and 3.9 ng/mL), 44 subjects with normal serum folate (between 4.0 and 17.9 ng/mL), and 38 subjects with high serum folate (above 18 ng/mL). Eighty-four percent of the subjects with subnormal serum folate and 56% of the subjects with low normal serum folate had more than 7.05 nmol/mL serum total homocyst(e)ine (ie, more than two standard deviations above the normal mean). Thirty-two percent of these subjects had more than a three-fold increase in serum total homocyst(e)ine. These observations support the hypothesis that depletion of tissue folate causes homocysteinemia in nonhomocystinuric subjects. Subnormal as well as low normal concentrations of serum folate appear to produce an accumulation of homocyst(e)ine. In addition, relatively normal levels of serum total homocyst(e)ine were observed in four pregnant women with low serum folate, supporting previous suggestions of an influence of female sex hormone(s) in homocysteine metabolism.

Hypertensive diseases of pregnancy and parity.

Chesley's classic long-term follow up study of eclamptic women clearly demonstrated the prognostic significance of parity in the differential diagnosis of various hypertensive diseases of pregnancy. Multiparous patients with eclampsia were different on long-term follow-up from primiparous eclamptic women. A logical conclusion from this observation is that multiparous and primiparous patient groups should be analyzed separately whenever hypertensive diseases of pregnancy are evaluated. This study is therefore an attempt to define the clinical profile of hypertension during pregnancy on the basis of parity alone. The prenatal and hospital records of 99 successive pregnant patients identified as hypertensive during a 1-year period at Mount Sinai Hospital Medical Center served as the study population. A group of 25 primiparous and 25 multiparous patients, all with uncomplicated pregnancies and normal deliveries in sequence during a 1-month period, served as normal nonhypertensive control groups. Various pregnancy parameters were statistically evaluated for study and control groups. Significant differences were found between hypertensive primiparous and multiparous patients in mean weight increase (p less than 0.05), gestational age at first increase in blood pressure (p less than 0.007), and time from first increase in blood pressure until delivery (p less than 0.008). The difference in birth weight was not significant between hypertensive groups. In contrast, among the control groups, multiparous patients had significantly larger offspring than primiparous patients (p less than 0.01). Weights of normal primiparous control women were almost identical to those of primiparous hypertensive women (3252.8 +/- 511.8 versus 3203.26 +/- 679.5 gm). Birth weights of offspring from multiparous hypertensive women, while not significantly different from those of primiparous hypertensive women, were significantly lower than those of multiparous control subjects (3093.94 +/- 898.7 versus 3593 +/- 305.6; p less than 0.01). No significant differences in a variety of laboratory findings, reflexes, edema, or mode of delivery were observed between the various groups. The clinical and laboratory presentation of hypertension in pregnancy may be similar between primiparous and multiparous patients. Nevertheless, clear differences do exist in both maternal presentation and impact of maternal disease on fetal growth and development. These differences strongly suggest a different pathophysiology as the underlying cause of hypertensive disease in primiparous and multiparous pregnant women.(ABSTRACT TRUNCATED AT 400 WORDS)

Protein-bound homocyst(e)ine. A possible risk factor for coronary artery disease.

The development of atherosclerotic changes and thromboembolism are common features in homocystinurics. Hence, we postulate a positive correlation between the level of homocyst(e)ine in the blood and the occurrence of coronary artery disease. Homocysteine is found either as free homocystine, cysteine-homocysteine mixed disulfide, or protein-bound homocyst(e)ine. In nonhomocystinuric subjects, most homocysteine molecules are detectable in the protein-bound form. Thus, protein-bound homocyst(e)ine in stored plasma which reflected total plasma homocyst(e)ine was determined in 241 patients with coronary artery disease (173 males and 68 females). The mean +/- SD total plasma homocyst(e)ine was 5.41 +/- 1.62 nmol/ml in male patients, 4.37 +/- 1.09 nmol/ml in male controls, 5.66 +/- 1.93 nmol/ml in female patients, and 4.16 +/- 1.62 nmol/ml in female controls. The differences between the patients with coronary artery disease and the controls were statistically significant (P less than 0.0005).

Type A score (Jenkins Activity Survey) and risk of recurrent coronary heart disease in the aspirin myocardial infarction study.

The Jenkins Activity Survey (JAS), a questionnaire developed to assess the type A behavior pattern, was administered to 2,314 participants in the Aspirin Myocardial Infarction Study. All had a myocardial infarction (MI) before entering the study and were followed for at least 3 years. The JAS type A score was not significantly related to risk of recurrent major coronary events (definite nonfatal MI and coronary death) in the group of 244 women, the group of 2,070 men, or the subgroup of 671 men who were employed full-time in professional, technical or managerial positions. These results indicate that the JAS type A score is not useful in assessing prognosis after MI. By inference, traits measured by the JAS type A score, such as competitiveness, orientation toward achievement and preference for a rapid pace of life, appear not to be associated with increased risk of recurrent major coronary events.

A randomized trial of adjuvant chemotherapy in head and neck cancer.

Ninety-five patients with squamous cell carcinoma of the head and neck were entered into a randomized study testing a two-week course of induction chemotherapy with methotrexate and leucovorin given prior to regional therapy. In addition, following regional therapy, patients randomized to chemotherapy were to receive similar methotrexate courses every three months for one year. Poor tolerance to this regimen after radiation and surgery led to a change in the chemotherapy following regional therapy to a combination of Adriamycin (Adria Laboratories, Columbus, Ohio) and cisplatin every three weeks for four cycles after the first 35 patients had been entered. Nine cases were ineligible and four lacked any follow-up data, leaving 82 analyzable cases. Using Cox regression analysis, no differences in the percentage of patients achieving disease control, the relapse-free survival, or the overall survival were identified between any treatment group. As has been described in many pilot studies of induction chemotherapy of head and neck cancer, chemotherapy responders had a more favorable disease-free survival than chemotherapy nonresponders in the total group of patients receiving adjuvant chemotherapy. However, correcting for imbalances in the expected three year disease-free survival of these patients, based on their disease site and stage, erased this difference, indicating tumor response to this regimen of chemotherapy is not an independent factor affecting disease outcome. The division of patients into arbitrary prognostic categories based on the expected outcome for each specific tumor site and stage proved to be a useful method for balancing treatment groups, given the multiple site-stage combinations within the upper aerodigestive tract. The defined prognostic categories were the single most sensitive predictors of relapse-free and overall survival.

A double-blind evaluation of sodium gradient hemodialysis.

In a double-blind, crossover trial, 7 chronic hemodialysis patients underwent three 4-week treatment periods. During one period, dialysate contained 135 mEq/l sodium. During another period, dialysate contained 143 mEq/l sodium. During the remaining period, we used "sodium gradient' dialysate, the sodium concentration of which was decreased from 160 to 133 mEq/l during each 4-hour dialysis session. Ultrafiltration was performed at a constant rate to achieve a predetermined post-dialysis weight. Interdialytic weight gain, thirst, blood pressure control, and incidence of side effects were monitored. There was a significant difference in interdialytic weight gain for the 3 treatments (p = 0.005). Interdialytic weight gain using 135 mEq/l sodium dialysate (2.2 +/- 0.9 kg, mean +/- SD) was significantly less than that using either 143 mEq/l sodium dialysate (2.6 +/- 0.8 kg) or sodium gradient dialysate (2.8 +/- 0.7 kg). Self-reported thirst tended to be less severe with 135 mEq/l sodium dialysate than with 143 mEq/l sodium dialysate or with sodium gradient dialysate, but changes in thirst were not statistically significant (p = 0.13). The incidence of intradialytic hypotensive episodes was comparable with the 3 levels of dialysate sodium. The results suggest that the described sodium gradient method does not prevent the increased interdialytic weight gain and thirst seen with other forms of high-sodium dialysis, and probably does not reduce the incidence of side effects.

Hemodynamic response to volume depletion in acutely uremic dogs.

Hemodynamic response to volume depletion by isolated ultrafiltration was compared in uremic (U) and nonuremic (N) conscious dogs. Fluid was removed at a constant rate until mean arterial pressure (MAP) decreased to less than 80 mmHg. Initial MAP was higher in the uremic dogs [132 +/- 8.6 (SD) mmHg] than in nonuremic controls (106 +/- 12, P less than 0.001). Initial cardiac index [U 4.97 +/- 0.831 X min-1 X m-2, N 4.44 +/- 0.62] and total peripheral vascular resistance index [(TPRI) U 2,160 +/- 353 dyn X s X cm-5 X m-2, N 1,976 +/- 420] were slightly, but not significantly, higher in uremic animals. Initial central venous pressure, wedge pressure, and plasma norepinephrine level were greater in the uremic dogs. At the end point of volume depletion, both uremic and nonuremic animals had achieved similar levels of TPRI, despite greatly attenuated or absent increases in plasma renin activity in the uremic group. At end point, blood volumes and plasma norepinephrine levels were comparable. The increase in pulse rate was higher in the uremic animals (59 +/- 37 pulses/min) compared with controls (25 +/- 52, P less than 0.05). In an additional group of uremic dogs, cardiovascular responses to hemorrhage and isolated ultrafiltration were compared and found to be similar, after allowance for blood viscosity changes had been made. The data suggest that in acutely uremic conscious dogs, despite reduced renin-angiotensin responses, hemodynamic adaptation to rapid volume depletion is not impaired.

Transdermal clonidine for hypertensive patients.

Seventeen moderately hypertensive patients, whose blood pressure was previously controlled with hydrochlorothiazide and oral clonidine (blood pressure 129 +/- 8/85 +/- 5 mmHg during therapy), were treated with a transdermal system involving application of one or more clonidine-containing patches (3 mg per patch) to the skin once a week. The patients continued to take 50 mg of oral hydrochlorothiazide daily. By four to eight weeks, 15 of 17 patients using the transdermal system had achieved baseline blood pressure levels (130 +/- 10/84 +/- 6 mmHg, NS). During the maintenance phase of transdermal therapy, plasma clonidine levels measured four hours (0.78 +/- 0.43 ng/ml), four days (0.89 +/- 0.48 ng/ml), and seven days (0.78 +/- 0.41 ng/ml) after patch application did not differ significantly from one another or from trough plasma clonidine levels (0.86 +/- 0.54 ng/ml) measured during oral clonidine therapy. The results suggest that, in moderately hypertensive patients, blood pressure can be controlled with a once-weekly application of clonidine-containing skin patches as effectively as with oral clonidine.

Epidemiologic aspects and factors related to survival in 384 Registry cases of clear cell adenocarcinoma of the vagina and cervix.

Three hundred and eight-four cases of clear cell adenocarcinoma of the vagina and cervix accessioned in the Registry as of December 31, 1978, have been analyzed. The annual incidence of these tumors has been found to correspond closely to the estimated usage of diethylstilbestrol (DES) for pregnancy support in the United States. The annual incidence of the DES-associated cases appears to have dropped in the United States in the past 2 years in comparison to 1973 to 1975. The risk of tumor development appears to be higher in young women exposed to DES early in intrauterine life than in those exposed later. The carcinomas are rare before the age of 14 years and an irregular peak in the age-incidence curve appears between 17 and 21 years followed by a decline. The 5-year survival rate is better for women over the age of 19 years than for younger patients, and a higher frequency oral contraceptive usage did not appear to influence the behavior of the tumor and the improved survival in those using this medication appears to be related to greater medical surveillance.

A dual-processor computer for arrhythmia analysis.

In order to facilitate the analysis of 24-hour tape recordings of ECG for rhythm by the AZTEC technique, we have configured a dual-processor system. It samples the ECG signal from a tape playback unit at a rate of 7501.88 Hz, allowing analysis at 30 x real time. An approach to error recognition and containment for the program, which utilizes the sampling of events, is discussed. Steps are taken to reduce errors caused by human sampling of "high risk" regions of the record.

Cytogenetics of methadone-managed and heroin-addicted pregnant women and their newborn infants.

Chromosomal damage was assessed prenatally and at delivery from 99 addicted pregnant women (80 from a methadone maintenance program and 19 heroin addicts) and their 101 off-spring at delivery. About 10% of the 27,907 cells scored showed chromosomal abberations. Chromosome damage was random, affected all chromosomes, and was mainly of the acentric fragment type. The per cent of hypodiploidy was significantly higher than the per cent of hyperdiploidy. In the mothers, no significant differences were found with respect to dosage and duration of methadone treatment and years of heroin abuse. No significant association was found between maternal variables and infant chromosome damage. Infants with low Apgar scores (1 to 6) had cells with significantly higher levels of chromosome damage than infants with higher scores (7 to 10).