Systematic review and meta-analysis.

In this review the usual methods applied in systematic reviews and meta-analyses are outlined. The ideal hypothesis for a systematic review should be generated by information not used later in meta-analyses. The selection of studies involves searching in web repertories, and more than one should be consulted. A manual search in the references of articles, editorials, reviews, etc. is mandatory. The selection of studies should be made by two investigators on an independent basis. Data collection on quality of the selected reports is needed, applying validated scales and including specific questions on the main biases which could have a negative impact upon the research question. Such collection also should be carried out by two researchers on an independent basis. The most common procedures for combining studies with binary outcomes are described (inverse of variance, Mantel-Haenszel, and Peto), illustrating how they can be done using Stata commands. Assessment of heterogeneity and publication bias is also illustrated with the same program.

Comparison of two procedures to estimate the hospital stay attributable to nosocomial infection: matched cohort study versus analysis of covariance of the total unmatched cohort.

We compare the results of a matched cohort study with those yielded by analysis of covariance (ANCOVA) in the cohort where the matched study was nested to assess whether the matching design underestimates (as it has been assumed) the extra length of hospitalization due to nosocomial infection. A total of 218 patients developed hospital infection in a cohort of 1483 general surgery patients; 161 were successfully matched 1:1 for surgical procedure, ASA score, age (+/-10 years), emergency-scheduled surgery, preoperative stay, and, whenever possible, number of diagnoses and sex. Unmatched infected patients (57, 23.1%) were different from matched ones. There were no differences for the variables between matched infected patients and their pairs. The matched cohort study overestimates the extra LOH due to hospital infection. The use of ANCOVA in the total cohort obviates the selection bias of the matched cohort design.

Epidemiology of surgical-site infections diagnosed after hospital discharge: a prospective cohort study.

OBJECTIVE: To study postoperative infections in hospital and after discharge, and to identify the risk factors for such infections. DESIGN: Prospective cohort study, with telephone follow-up for 1 month after hospital discharge. SETTING: The general surgery service of a tertiary hospital in Spain. MAIN OUTCOME MEASURE: In-hospital and postdischarge surgical-site infection (SSI), always confirmed by a physician. RESULTS: Of the 1,506 patients initially enrolled, 29 died during hospital stay, and 33 were lost to postdischarge follow-up. An SSI was identified prior to discharge in 123 patients and after discharge in 103. For several variables (age, serum albumin, glycemia, lengths of preoperative and postoperative hospital stay, etc), there were no differences between patients with postdischarge SSI and noninfected patients; however, there were differences detected between patients with postdischarge SSI and in-hospital SSI, as well as between patients with in-hospital SSI and noninfected patients. The analysis of risk factors showed that most predictors for in-hospital SSI did not behave in the same manner for postdischarge SSI. Stepwise logistic regression only identified chemoprophylaxis, age (advanced age was a preventive factor), and body mass index as independent risk factors for postdischarge SSI. Differences in risk factors between in-hospital and post-discharge SSIs remained even after controlling for time from operation to diagnosis. CONCLUSIONS: Most predictors of in-hospital SSI were not predictors of postdischarge SSI.

Total cholesterol, HDL-cholesterol, and risk of nosocomial infection: a prospective study in surgical patients.

OBJECTIVE: To study the relationship between serum high-density lipoprotein cholesterol (HDL-C), total serum cholesterol, and nosocomial infection in patients undergoing general surgery. DESIGN: Prospective cohort study, with an extended follow-up to 1 month after hospital discharge. SETTING: The general surgery service of a tertiary hospital. MAIN OUTCOME MEASURE: Nosocomial infection, mainly surgical-site infection (SSI), urinary tract infection, respiratory tract infection (RTI), and bacteremia. PATIENTS: 1,267 surgery patients aged 10 to 92 years. RESULTS: 182 subjects acquired 194 nosocomial infections, a cumulative incidence of 14.5%; most (116, 62.3%) were postoperative wound infections. There was an increase in infection risk at low levels of HDL-C, and both low and high total cholesterol levels. After adjusting simultaneously for several confounders, including total cholesterol, low levels of HDL-C (< or = 20 mg/dL) yielded an odds ratio (OR) of 2.2 (95% confidence interval [CI95], 0.6-7.9) for SSI and an OR of 10.3 (CI95, 0.7-151.5) for RTI. Otherwise, no trend was observed between HDL-C levels and infection risk, and no increased risk of nosocomial infection was observed for HDL-C values in the range of 21 to 49 mg/dL. Serum cholesterol showed a U-shaped relationship with nosocomial infection risk. Both low levels (below 102 mg/dL) and high levels (above 290 mg/dL) of total cholesterol were associated with a higher risk of SSI (mainly those caused by gram-negative bacteria) and RTI in comparison with the reference group (139-261 mg/dL). CONCLUSIONS: Serum HDL-C and total cholesterol seem to be associated with the risk of nosocomial infection in surgical patients.

Nosocomial infections in surgical patients: comparison of two measures of intrinsic patient risk.

OBJECTIVE: To compare, in subjects undergoing general surgery, two measures of intrinsic patients risk for nosocomial infection: the Study on the Efficacy of Nosocomial Infection Control (SENIC) index and the National Nosocomial Infection Surveillance (NNIS) System index. DESIGN: Prospective cohort study, with follow-up for 1 month after hospital discharge. SETTING: The general surgery service of a tertiary hospital. MAIN OUTCOME MEASURE: Surgical-site infection. PATIENTS: 1,483 subjects aged 10 to 92 years. RESULTS: During follow-up, 155 patients developed nosocomial infection, yielding a cumulative incidence of 10.5%. The NNIS index showed a linear trend with both crude and adjusted (for SENIC index) rates of surgical-wound infection. The SENIC index did not exhibit any linear trend with adjusted (for NNIS index) rates of surgical-wound infection. To delineate whether the SENIC index added explanatory information to the NNIS index (or vice versa), we regressed each variable on the other. Logistic regression analyses confirmed the results of stratified analysis: residuals of the NNIS index added discriminating ability to the SENIC index, whereas residuals of the SENIC index did not improve the predictive power of the NNIS index. CONCLUSIONS: The NNIS index had a better ability than the SENIC index for discriminating and predicting risk of surgical-wound infection.

The effect of frequency of chart review on the sensitivity of nosocomial infection surveillance in general surgery.

OBJECTIVE: To assess the sensitivity of different frequencies of nosocomial infection surveillance (NIS) in general surgery. DESIGN: Data obtained with a prospective daily NIS are compared with those of hypothetical cross-sectional studies carried out with different frequencies (from one weekly visit up to one visit every other day). SETTING: General surgery services at three hospitals. MAIN OUTCOME MEASURE: Sensitivity in the detection of nosocomial infection (overall and stratified by site), compared to a gold standard of prospective surveillance of every patient's complete medical record daily from the first day after surgery until discharge and once more after discharge. PATIENTS: 5,859 patients. RESULTS: 837 nosocomial infections were detected by the gold standard (58.8% were surgical-site infections [SSI]). The sensitivity of weekly NIS for all infections was 74.5% (95% confidence interval [CI95], 71.4%-77.5%) and varied from 65.1% (CI95, 56.2%-73.3%) for urinary tract infection to 83.3% (CI95, 62.6%-95.3%) for respiratory tract infection; it was 76.4% (CI95, 72.4%-80.1%) for SSI. As expected, sensitivity increased with the frequency of NIS. Performing NIS every 4 days improved sensitivity significantly, to 82.3% (CI95, 79.5%-84.8%) for all infections and 83.3% (CI95, 79.7%-86.5%) for SSI. One visit every other day increased the sensitivity for all infections by another 4.9%, mainly due to increased detection of urinary tract and other less severe infections. CONCLUSIONS: The sensitivity of two visits a week exceeded that of one weekly visit by approximately 8%, and one visit every other day added another 5% increase. Results varied according to duration of infection and postdischarge hospital stay.

Usefulness of intrinsic infection risk indexes as predictors of in-hospital death.

OBJECTIVE: Comparison of two measures of intrinsic infection risk for predicting in-hospital mortality risks among subjects undergoing general surgery: the Study on the Efficacy of Nosocomial Infection Control (SENIC) index and the National Nosocomial Infection Surveillance (NNIS) index. DESIGN: Prospective cohort study on 1483 patients admitted to the service of general surgery of a tertiary hospital. The main outcome measure was in-hospital death. Relative risks, crude and multiple-risk factor adjusted for by logistic regression analysis, and their 95% CIs were estimated. RESULTS: During follow-up, 33 patients (2.2%) died. Both the SENIC and the NNIS indexes appeared related to in-hospital mortality risk in crude data. After several confounders (age, sex, severity of illness, American Society of Anesthesiologists score, serum creatinine, serum albumin, stay at the intensive care unit, length of operation, type of surgical wound, and preoperative stay) were controlled for, the SENIC index showed a borderline significant trend with mortality (p = 0.052), whereas the trend was significant for the NNIS index (p = 0.026). The NNIS index also showed a linear trend with both crude and adjusted for (SENIC index) risk of death. The SENIC index did not exhibit any linear trend with adjusted for (NNIS index) risk of surgical wound infection. To delineate whether the SENIC index added explanatory information to the NNIS index (or vice versa), we regressed the SENIC index on the NNIS index (and vice versa) and computed a set of residuals for both indexes. In logistic regression analyses, the residuals of NNIS index added meaningful information to the SENIC index, whereas the residuals of the SENIC index did not add any relevant information to the NNIS index. These results remained unchanged after controlling for several confounders. CONCLUSIONS: Both the SENIC and the NNIS indexes are good predictors of in-hospital mortality risk. The NNIS index had greater capability for discriminating and predicting risk of dealth.

Surgical wound infections diagnosed after discharge from hospital: epidemiologic differences with in-hospital infections.

BACKGROUND: The purpose of this study was to study postoperative infections detected in hospital and after discharge and to identify risk factors for such infections. METHODS: A prospective cohort study was used, with follow-up of 30 days after hospital discharge, on 1483 patients admitted to the general surgery service of a tertiary care hospital. The main outcome measure was surgical wound infection (SWI). Relative risks, crude and multiple risk factors adjusted for by logistic regression analysis, and their 95% confidence intervals (CIs) were estimated. RESULTS: During follow-up 155 patients showed evidence of nosocomial infection, 134 in hospital and 21 at home, yielding a cumulative incidence of 10.5%. According to several variables (age, American Society of Anesthesiologists score, serum albumin, the SENIC and National Nosocomial Infections Surveillance indexes of intrinsic patient risk, length of hospital stay, etc.) there were no differences between patients with postdischarge SWI and uninfected patients; however, differences were detected between postdischarge SWI and in-hospital SWI, as well as between patients with in-hospital SWI and patients without infections. The analysis of risk factors showed that most predictors for in-hospital SWI did not behave in the same manner for postdischarge SWI. Stepwise logistic regression analysis identified cancer (odds ratio = 4.5, 95% CI = 1.7 to 12.2, p = 0.003) and surgeon performing the operation (for medium risk, OR = 4.4, 95% CI = 0.9 to 21.3, p = 0.059; for high risk, OR = 3.0, 95% CI = 0.7-13.3, p = 0.144) as independent risk factors for postdischarge SWI. CONCLUSIONS: There were important epidemiologic differences between in-hospital SWI and postdischarge SWI; most risk factors for in-hospital SWI are not predictors for postdischarge SWI.

Efficacy of surveillance in nosocomial infection control in a surgical service.

OBJECTIVE: The purpose of this study was to assess the efficacy of surveillance of nosocomial infection in infection control at a service of general surgery. DESIGN: A surveillance study that included 1483 patients with a prospective identification of nosocomial infection was carried out. Its results were discussed with the staff, and a program on nosocomial infection control was implemented. One year after the pre-intervention study, a similar study that included 1506 patients was done. The main outcome measure was nosocomial infection. Incidence rates, incidence rate ratios, crude and multiple-risk factor adjusted for by Poisson regression analysis, and their 95% confidence interval rates were estimated. RESULTS: The characteristics of the patients enrolled in both studies were compared. After the intervention, the trend was to attend patients with more severe conditions: higher frequency of liver failure, chronic obstructive lung disease, higher proportion of dirty surgical wounds, and higher scores of both Study on the Efficacy of Nosocomial Infection Control (SENIC) and National Nosocomial Infections Surveillance indices. There were no significant differences in emergency surgery, duration of surgery, age, and sex. After the intervention, unnecessary chemoprophylaxis was drastically reduced, and a significant reduction in preoperative stay was observed. The nosocomial incidence rate fell from 18.4 to 14 per 1000 patient-days. This reduction yielded an incidence rate ratio of 0.56 (95% confidence interval, 0.43%-0.74%) adjusted for several variables (SENIC index, serum creatinine level, serum albumin level, antihistamine H2 level, surgical wound, body mass index, chemoprophylaxis, and community-acquired infection). Significant reductions in surgical site infection and urinary tract infection were observed, but the rate of respiratory tract infection remained unchanged. CONCLUSIONS: Surveillance was effective in reducing nosocomial infection.

Menopausal hormone replacement therapy and breast cancer: a meta-analysis.

A meta-analysis was performed to determine whether the scientific literature provides enough evidence that hormone replacement therapy after menopause increases the risk of breast cancer. Studies were located by MEDLINE, supplemented by a hand search of all the references in the articles located. The papers were graded as to quality. Those considered unbiased were combined using Woolf's method. Thirty-seven original studies were found: 23 case-control, 13 cohort, and one clinical trial. Overall, a small but statistically significant relative risk (RR) figure of 1.06 was calculated. Women who experienced natural menopause seemed to be at increased risk (RR = 1.13). A significant weighted RR was observed in current hormone replacement therapy users, especially in those who had natural menopause (RR = 1.63). A nonsignificant increasing trend was found between duration of hormone replacement therapy and breast cancer risk, although the opposite was seen when the association was analyzed by time since last use. These results imply that hormone replacement therapy could promote breast cancer.

Usefulness of intrinsic surgical wound infection risk indices as predictors of postoperative pneumonia risk.

The main objective of this report was to use two indices of intrinsic surgical wound infection risk, the SENIC index (Haley et al., 1985) and the NNIS index (Culver et al., 1991), to predict risk of postoperative pneumonia in general surgery patients. A prospective cohort study on 1483 patients admitted under the general surgery speciality of a tertiary hospital was performed. The main outcome measure was postoperative pneumonia. Relative risk and their 95% confidence intervals (CIs) were estimated. Stepwise logistic regression analysis was used to select the main determinant predictors. During follow-up, 19 (1.3%) patients acquired postoperative pneumonia. Common risk factors of postoperative pneumonia were identified: mechanical ventilation, age, upper abdominal surgery, severity of illness, obesity, hypoalbuminaemia, and use of histamine type 2 receptor antagonists. Both the SENIC and the NNIS indices showed a statistically significant association (P < 0.001) with postoperative pneumonia risk: the higher the score the greater the risk. Stepwise logistic regression analysis selected five variables: (1) mechanical ventilation [odds ratio (OR) = 9.8, 95% CI 2.7-35.6]; (2) upper abdominal surgery (OR = 4.7, 95% CI 1.6-13.9); (3) chronic lung disease (OR = 5.9, 95% CI 1.7-21.2); (4) the NNIS index (OR for each point = 2.2, 95% CI 1.1-4.4); and (5) obesity, measured by a body mass index greater than the 90th percentile (OR = 2.9, 95% CI 0.9-9.4). In conclusion, both the SENIC and the NNIS indices were related to postoperative pneumonia risk. The NNIS index may be a better predictor.

Nosocomial infection, indices of intrinsic infection risk, and in-hospital mortality in general surgery.

The objectives of this paper are to assess whether two indices of intrinsic infection risk (the SENIC and the NNIS index) predict in-hospital mortality and the attributable in-hospital mortality due to nosocomial infection in surgical patients. A prospective study on 4714 patients admitted to three hospitals has been carried out. The relative risk and its 95% confidence interval (CI) were estimated. Multiple-risk factors adjusted for odds ratios (OR) were yielded by logistic regression analysis. Overall, 119 patients (2.5%) died before hospital discharge. Both the SENIC and the NNIS indices were related to in-hospital mortality in crude data. After controlling for several variables (age, sex, ASA score, cancer, renal failure, diabetes mellitus, stay at the ICU), the SENIC index did not show any significant trend with mortality (P = 0.252), whereas the trend was significant for the NNIS index (P < 0.001). Risk of death in patients with one nosocomial infection was 7.5%, and in patients developing more than one nosocomial infection was 17.1%. After adjusting for several confounding variables, the development of an organ/space surgical site infection was significantly related to mortality (OR = 4.5, 95% CI 1.5-15.6) as was blood infection (OR = 17.3, 95% CI 3.5-87.0). The association of a surgical site infection and either a respiratory tract infection or a blood infection also increased significantly the risk of in-hospital mortality (OR = 3.3, 95% CI 1.2-8.7). In conclusion, the NNIS index is a good predictor of in-hospital mortality. Patients developing an organ/space surgical site infection and/or a blood infection have an increased risk of in-hospital mortality.

Oral contraceptives and breast cancer. A meta-analysis.

Since the relationship between oral contraceptive use and breast cancer has not been consistent we undertook a meta-analysis of studies published to date. Papers were located by searching the MEDLINE data base, supplemented by a hand search of all the references in the articles recovered. Studies were graded as to quality. Those judged as probably unbiased were included in the analysis. The method of Woolf was used to combine relative risks. Forty-seven studies were collected: 40 case-control and 7 cohort studies. Thirty-nine of these were considered unbiased. The main results observed were: RR was 1.06 (1.02-1.10) for all studies and 1.14 for premenopausal cancer. For premenopausal cancer, higher RRs were observed in women who early used oral contraceptives with a significant linear dose-response effect: 1.25 (1.10-1.44) in OC users before age 25, and 1.17 (1.06-1.30) in users before the first full-term pregnancy. We conclude oral contraceptive use may be a risk factor for premenopausal breast cancer. Limitations to our research are discussed.

PIP: 40 case control and 7 cohort studies of the link between oral contraceptives and breast cancer were combined in a meta-analysis by the method of Woolf to estimate overall relative risks. Articles were found by a Medline search from January 1966-June 1990, a hand search of the references, and a journal search for 1990. Papers were graded for bias by 2 blinded readers. The analysis of the 39 studies considered unbiased generated a global relative risk of 1.06, and an increased risk of 1.06, and an increased risk of 1.14 for premenopausal breast cancer. For premenopausal cancer, risks were 1.25 for use before age 24, and 1.17 for use before the 1st term pregnancy. A significant linear dose response was seen, in terms of months of exposure. There was an indication of latency, since risk estimates were higher from data collected after 1981. There was no evidence of a relationship between pill use and parity, family history, or history of benign breast disease. The dad are consistent with the theory that oral contraceptives can be acting as cancer initiators or promoters. There was not enough data to judge whether users of current lower dose pills have lower risks for premenopausal breast cancer. Future studies should address the difference between premenopausal and postmenopausal cancer.

[Biases in meta-analysis]. / Sesgos en el metaanálisis.

The objective of the present article is to review the biases in meta-analysis. This procedure can be included within observational studies, and its biases follow the same classification: selection, information and confounding bias. Selection bias can always be present because of the impossibility of enumerating the reference population. More interest deserves publication bias, influenced by many facts: significant results, design type, language of publication, the existence of conflicts of interest, sample size, institutional prestige, existence of prejudices, and regression to the mean. There are several graphic and quantitative procedures to detect the error under certain assumptions. The correction of this bias is quite unlikely. It is easier to guarantee that the selection process of the studies among those located is not subjected to errors, if structured protocols are used. Information bias has been less studied, but it has also been documented. Control of confounding bias depends on the identification of this error in the individual studies, but it can also be introduced through certain analytic techniques, as the ecologic fallacy. Lastly, it is recommended not to use the procedure of Peto in the analysis as this technique provides biased results.

[Incidence of nosocomial infection surveillance in general surgery]. / Frecuencia de la vigilancia de la infección nosocomial en cirugía general.

BACKGROUND: We have not found any study assessing the sensitivity of different frequencies of nosocomial infection surveillance (NIS). The objective of this study is to contribute to the ascertainment of the frequency of the NIS surveillance in surgical patients, where NIS has shown to be most efficient. METHODS: A prospective cohort study on 1,483 patients undergoing general surgery in a reference hospital was performed between November 1992 and July 1994. A daily NIS, plus review of clinical chart after hospital discharge, performed by two trained clinicians was taken as gold standard. The expected number of infections detected according to different schedules of NIS (from a minimum of weekly examination to a NIS performed every other day) was estimated taking into account the duration of infection and the hospital stay after infection healing. This number was corrected multiplying it by 0.9 (sensitivity of NIS without reviewing clinical chart after discharge). Confidence intervals (CI) were estimated using the binomial's exact method. RESULTS: 235 nosocomial infections were detected by the gold standard. The sensitivity of the weekly NIS for all infections was 78.3% (95% CI = 72.5-83.4); it varied from 63.6% (95% CI = 45.1-79.6) for urinary tract infections (UTI) to 85% (95% CI = 62.1-96.8) for respiratory tract infections; for surgical wound infections, it was 80.1% (95% CI = 72.4-86.5). As expected sensitivity increased with the frequency of NIS. A frequency of NIS every four days yielded a sensitivity for all infections of 86.4% (IC 95% = 81.3-90.5), 78.8% (CI 95% = 61.1-91.0) for UTI and 86.8% (CI 95% = 79.9-92.0) for SWI. CONCLUSIONS: Most nosocomial infections were detected by one visit a week. Twice visits a week improved the sensitivity of NIS in about 10%. It is important to emphasize that results change according to duration of infection and post-discharge hospital stay.

[Inclusion of research quality in meta-analyses]. / Inclusión de la calidad de una investigación en el metanálisis.

The objective of this review is how study quality is taken into account in the synthesis of information from several studies. Meta-analysis has been criticized for failing into account differential quality of evidence in primary studies, allowing poor studies to drive out good science by weight of numbers. Notwithstanding there is presently no agreement on the inclusion of study quality score in quantitative meta-analysis; on the other hand, agreement does exist on knowing study quality. However, and although there is a certain level of consensus about what a good study is and on what characteristics it should have, no evaluation protocol on study quality is uniformly accepted. Furthermore, there are no data on the reliability and validity of quality questionnaires. There are several procedures to include study quality in meta-analysis: a) it is generally accepted that bias-free studies should be pooled only; b) most researches reject to include quality scores in the weight; and c) no agreement is reached on the use the quality scores as a covariate in graphical methods and regression analysis. It may be superfluous or misleading.

[Proportional mortality studies: criteria for the selection of participant groups]. / Estudios de mortalidad proporcional: criterios de elección de los grupos participantes.

Proportional mortality designs are used widespread in occupational epidemiology. In this review those biases which can affect them, mainly the healthy worker bias, are discussed. Several options for their analysis and the assumptions to be accomplished for validity are reviewed: proportionate mortality analysis, standardized mortality ratio, and odds ratio. The inclusion of dead participants in a research exhibits several drawbacks. Starting out from the analysis of this sort of designs, the first criterium to select diseases is similar to case-control studies: the reference group must not include diseases related with the exposure under study. Analyzing the relationship between mortality and incidence rates, criteria to select diseases to be investigated by proportional mortality studies are offered. These designs yield a valid inference when the disease is rare and irreversible. If the exposure shortens duration of disease, a toward-the-null bias is introduced. The direction of bias is variable under other circumstances, although it shows a trend to be negative.

[Meta-analysis in epidemiology (1): the general characteristics]. / Metaanálisis en epidemiologia (primera parte): caracteristicas generales.

In the present work the general characteristics, aims and applications of meta-analysis in public health are described. The general rules and the relevant steps for the development of this type of studies are considered. These points are illustrated with a number of real examples and the advantages and limitations of the different methodological options are commented.

[Meta-analysis in epidemiology (2): quantitative methods]. / Metaanálisis en epidemiología (Segunda parte): métodos cuantitativos.

The main goal of this paper is to offer an introduction to the quantitative methods of meta-analysis used in epidemiology/public health, emphasizing on the most widespread used. We give also a brief reference of the methods to recognize publication bias. The limitations and advantages of the methods commented on are discussed.

[Prevalence of mental subnormality recorded in the province of Jaen]. / Prevalencia de la subnormalidad mental censada en la provincia de Jaén.

It is done a cross-sectional study of the prevalence of mental retardation in the province of Jaén (Spain), based on the cumulated cases reported to the Institute of Social Services of the Social Security (INSERSO) until January 1, 1984. The main objective of this research is to know the prevalence of mental retardation in our province, and to analize its relationship with some sociodemographic variables. The prevalence obtained has been 4.09%, being a 59.27% of the prevalence due to endogenous/unexplained etiology. Mild subnormality constitutes a 8.4% of cases, being closely related with the size of the locality (municipio). Prevalence of mental subnormality showed a significative linear association with the size of the locality (r = -0.904, p less than 0.05), and with the population growth (r = -0.929, p less than 0.01). This relationship was not observed neither with the level of per capita income nor the altitude of the locality. A multiple regression analysis is made for every etiology of mental retardation and several social and geographic variables. We conclude that the figures for mental retardation in our province show a direct relationship with local development parameters.