From Wikipedia

Contents 1 Methods 1.1 Ethics 1.2 Design 1.3 Database 1.3.1 Overall description 1.3.2 Inclusion and exclusion 1.4 Variables 1.4.1 Outcomes 1.4.2 Predictors 1.4.3 Confounders 1.5 Data Analysis 1.5.1 Statistical methods 1.5.2 Missing values 1.5.3 Statistical software 2 Results 2.1 Results from inclusion and exclusion criteria 2.2 Sample description 2.3 Bivariate analysis 2.4 Models

Methods

Ethics

Enter the information regarding IRB (Institutional Review Board) approval for your project.

• Approval from the Duke Institutional Review Board was obtained prior to the initiation of this project.
• Ethical approval for this study was obtained from the Addenbrooke's hospital research ethics committee.

Design

Describe the design used in the study (For example, prospective, retrospective, case control, etc.)

• We performed a secondary analysis of a national administrative database.

Database

Overall description

You should describe the database that provided you with the cases that will be used in the study. The description of the database should include following:

• during what time period the data were collected
• which institution or institutions are being utilized
• how the registration process was implemented
• how the cases were recruited
• how the data was entered
• who entered the data
• how the data was stored
• was there any quality control checks performed on the data

Also, mention any other sources of data (for example case notes, hospital records) that you might use for your study

In this paragraph you should describe the time period in which the study was carried out, the number of patients initially evaluated, places where the study took place, and how patients were recruited.

• The 1997 Healthcare Cost and Utilization Project Nationwide Inpatient Sample (HCUP-NIS) database, Release 6 8 , was used for this study. This database was developed by the United States Agency for Healthcare Research and Quality as part of a partnership between industry, federal, and state-level agencies to help track and analyze trends in health-care utilization, cost, quality, and outcomes. Hospitals were selected for the HCUP-NIS database according to the American Hospital Association (AHA) definition of community hospitals as all nonfederal, short-term general and specialty hospitals, excluding hospital units of institutions. The 1997 HCUP-NIS database Release 6 includes HCUP and AHA hospital identifiers, synthetic provider and patient identifiers, diagnosis-related groups (DRG) version 10, and procedure and diagnostic codes classified according to the International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM). Patient-specific demographic data and hospitalization information coded by synthetic identifiers are also available.

• The setting for our study was North Carolina, whose population is almost equally divided between urban and rural residents. Twenty-two percent of the population is black.7 Almost 600 chiropractors practice in North Carolina, and previous research by our group has demonstrated that 39 percent of persons seeking care for acute back pain go to a chiropractor first.Practitioners:Using medical and chiropractic state-licensure files, we randomly selected practitioners from six strata: urban primary care physicians, rural primary care physicians, urban chiropractors, rural chiropractors, orthopedic and neurologic surgeons, and primary care physicians and a small number of nurse practitioners and physician's assistants at a group-model health maintenance organization (HMO). We defined primary care as family practice, general internal medicine, or general practice. Very few osteopathic physicians practice in North Carolina. None of the neurologic surgeons who were selected saw a substantial number of patients with acute low back pain. Since few orthopedic surgeons practice in rural areas, we did not divide this group into rural and urban practitioners.The practitioners were aware of the overall purpose of the study but not of the specific outcome or utilization variables. On average, 74 percent of the eligible practitioners invited to participate in the study agreed to do so, ranging from 65 percent of the primary care providers to 87 percent of the HMO and orthopedic providers. A total of 208 practitioners participated in the study: 39 urban primary care practitioners, 48 rural primary care practitioners, 32 urban chiropractors, 32 rural chiropractors, 29 orthopedic surgeons, and 28 physicians and nurse practitioners or physician's assistants in a group-model HMO.Patients:The practitioners invited consecutive patients with acute low back pain to participate in our study. The practitioners obtained consent from the patients and recorded basic information from the history and physical examination at the initial office visit. Staff members of the University of North Carolina Survey Research Unit performed all interviews. The study was observational, and we made no attempt to influence the practitioners' decisions about diagnostic tests or treatments. Staff members in each practice kept a list of patients recruited for the study, allowing an assessment of approximate recruitment rates. Fifty percent of patients with back pain seen by the practitioners were eligible for the study, and only 8 percent of those who were eligible declined enrollment. The main reasons for ineligibility were chronic pain and previous treatment for the current episode of pain. Patients were paid $20 for the time they spent answering interview questions. They were told that the purpose of the study was to determine how long back pain usually lasts and the types of treatments used. Members of the Survey Research Unit contacted the patients by telephone shortly after the index office visit. The median time from the index visit to the base-line telephone interview was seven days.(Carey, 1995).

Inclusion and exclusion

Here you should provide a description of the sample and how it was obtained. Describe the inclusion and exclusion criteria. A flow chart (example flow chart) might be used to make your inclusion and exclusion more explicit. The combination of inclusion and exclusion criteria should aim at making the study sample as homogeneous as possible. Although a flow chart can be used, you should not present socio-demographic results in this section -- keep it for the Results section. Provide as much detail about each of the inclusion and exclusion criteria as possible, so that readers can both reproduce your sample as well as determine how closely your sample compares to their patient population.

• Paraffin-embedded formalin-fixed liver tissue was obtained from the explant tissue of 69 patients who underwent orthotopic liver transplantation for HCC between 1985 and 2001.

• The 1997 NIS approximates a 20% stratified sample representative of community hospitals in the United States. Sampling strata were used to create the NIS based on 5 hospital characteristics to ensure maximal representativeness of the US population (geographic region, ownership, location, teaching status, number of beds). Thus, the results of the present analysis can be extrapolated to the entire US patient population undergoing LA and OA in the US during 1997.

• Patients were included in the present analysis if they were female and had either BCT (ICD-9 codes 85.21, 85.22, 85.23) or BAT (ICD-9 codes 85.41, 85.43) for breast cancer (ICD-9 codes 174.0, 174.1, 174.2, 174.3, 174.4, 174.5, 174.6, 174.8, 174.9). Breast cancer patients who underwent bilateral breast cancer surgery (ICD-9 codes 85.42, 85.44, 85.46, 85.48, or two of the following procedure codes occurring in any combination: 85.21, 85.22, 85.23), radical mastectomy (ICD-9 code 85.45), or extended radical mastectomy (ICD-9 code 85.47) were excluded from our investigation. Also, patients with metastatic disease (except to regional lymph nodes: ICD-9 code 196.3) were excluded. The exclusion criteria were applied to impart homogeneity to our patient population. We performed stratified analyses based on whether BCT or BAT was performed.

• Practitioners were eligible to participate in the study if they provided ambulatory care more than half the time and saw patients with acute low back pain who had not been referred by other providers. We did not include physical therapists as primary care givers for patients with acute back pain, since such patients rarely seek care from a physical therapist first.The criteria for enrollment included back pain of less than 10 weeks' duration, no previous care received for the pain, no history of back surgery or cancer, and no pregnancy at the time of the initial visit. In addition, the patient had to have a telephone and speak English.(Carey, 1995).

Variables

The variables should be described such that there is enough detail for each one of them and the reader can know details about categorization of continuous variables and units of measurement. For variables that imply a specific measurement tool (e.g., ultrasound, MRI, etc), you should provide details about the tool that was used to perform the measurements (e.g., ultrassound machine, resonance scanner, etc). For variables where there is potential disagreement among raters, you should describe, whenever possible, how a study on inter and intra-observer reliability was conducted. If the measurements require a certain degree of expertise by the measurer, you should provide information about the background and experience of that person or persons performing the measures. The description should specify the unit used to measure the event. In your study seek to identify and describe variables that belong to the one of the categories listed below.

Outcomes

Outcomes are the endpoints of your study. For example, for a study trying to evaluate patient survival and hepatocellular carcinoma, a variable indicating whether the subject died or not, is an outcome. They correspond to the variables that indicate results, in other words they are the possible results for situation in question. Examples of this of type of variable include: deceased, (which can be expressed as yes or no), cure (yes or no), recurrence of tumor (yes or no), complications related to procedures. In this section you should not provide results about each of the outcomes, but provide details on how each outcome was measured.

• Length of Hospital Stay. The length of hospital stay (measured in days) is defined as the difference between date of admission and date of discharge of the patient. Length of stay was coded as 0 for patients discharged during the day of admission.

• Outcomes considered in the study included in-hospital mortality rates, in-hospital postoperative complications, and length of stay in the hospital (see Appendix). Postoperative complications included pulmonary embolus, deep venous thrombosis in the lower extremity, and postoperative wound infection. Each complication was identified according to the ICD-9-CM codes (see Appendix).

• The primary outcome was the date of return to a functional status equivalent to that before the onset of pain.We also assessed the patient's work status at each interview. Satisfaction with care, determined with the use of a questionnaire designed for a previous study of back pain was assessed at the interview during which the patient reported complete recovery or at 24 weeks, if the patient had not previously reported complete recovery.Charges for medication were calculated as the average wholesale cost to the pharmacist plus 40 percent.(Carey, 1995).

Predictors

Predictors are, as the name says, the main hypothesized predictors of outcome. It is a class of variables that does not describe the results as yes (as the outcomes), but yes factors that can have direct influence on the result. Some examples are: type of treatment, number of surgeries done by a surgeon/year, surgeon years of experience (predicting the best surgical results), stage of the tumor (how advanced, worst results). For example, in a trial comparing two drugs and their impact on subject survival, the variable indicating which drug was given to each subject will be the main predictor. In this section you should not discuss how the variables were analyzed, simply describe the variables as they were measured.

• The main effects of interest in this investigation were those of hospital volume and surgeon volume. The hospital volume, which was the total number of primary and revision total knee arthroplasties performed in each hospital or hospital system, was divided into four groups: fewer than eighty-five procedures, eighty-five to 149 procedures, 150 to 249 procedures, and 250 procedures or more. The synthetic primary surgeon number taken from the HCUP-NIS data field was used to determine the total number of primary and revision total knee arthroplasties performed by a surgeon in 1997. Surgeon volume was also divided into four groups: fewer than fifteen procedures, fifteen to twenty-nine procedures, thirty to fifty-nine procedures, and sixty procedures or more. The cutoff values for the volume groups were based on the use of regression splines during exploratory analysis and attaining approximately equal percentages of patients in each category. Primary surgeon identifiers were available for approximately 52% of the primary and revision total knee arthroplasties. When surgeon identifiers were missing, physician identifiers were used as substitutes. Unique synthetic surgeon and physician identifiers represent two separate variables and are available in the HCUP-NIS database to indicate the care providers of the patient. The substitution of physician identifiers for missing surgeon identifiers is a valid method as the concordance between primary surgeon and physician identifiers is 97.5% in our database. Moreover, missing surgeon volumes were imputed with use of expectation maximization algorithms, with the final estimates for odds ratio having a maximum variation of only 6.3%. This slight variation did not significantly affect any of the associations found in our study.

• The samples were assessed for tumour grade in accordance with the World Health Organisation classification, where 1, well differentiated; 2, moderately differentiated and 3, poorly differentiated. Since it remains to be determined what is the most important predictor of outcome, the predominant grade of tumour or the area with the worst grade the most were both recorded. Tumour necrosis, presence or absence of microvascular invasion (measured according to the classification of Paget et al.) and tumour size (measured in mm) were also recorded. Where tumours were multiple, all tumours were assessed. In these cases the tumour with the worst histological parameters was utilised for the analysis.

• Information on demographic characteristics, use of health care services, and functional status was collected at the time of the base-line interview and at 2, 4, 8, 12, and 24 weeks or until the patients declared themselves "completely better." (Carey,1995).

Confounders

Confounders are any variables that can affect the relationship between outcomes and predictors. We need to pay strict attention to these variables in order to avoid indirect influence and undesired results. Examples of these variables include: age, comorbities, history of previous treatment of the tumor, severity of the patients illness before they undergo surgery. For example, age is a confounding factor in the association between type of treatment for ACL reconstruction and postoperative level of functional activity, because a group of older aged patients may have less favorable post-operative results regardless of the type of procedure they underwent. It is not necessary to justify your inclusion of specific confounders, unless the choice is unusual.

• Patient-specific covariates include age, gender, race, type of admission, median income, comorbidity (according to the Charlson index as modified by Deyo et al.), arthritis diagnosis, and patient disposition. The Charlson index modified by Deyo et al. summarizes comorbidity with an index score that provides a single parameter for measuring numerous comorbidities. Each patient diagnoses, as defined by the ICD-9-CM diagnosis codes, were assigned a value of 1, 2, 3, or 6 as a representation of the severity of illness. The patient composite score was then determined by adding up the respective ICD-9-CM code values. Hospital-specific covariates include size in terms of number of beds, region, ownership, and teaching status.

• A Cox proportional-hazards model was used to estimate the time to functional recovery, with adjustment for covariates that might confound the relation between the type of practitioner and functional recovery.(Carey, 1995).

Data Analysis

Statistical methods

Describe the statistical methods, including the methods used for control of confounding. If applicable, describe the methods used for subgroup analysis and sensitivity analysis.

• Descriptive statistics included means or percentages and 95% confidence intervals. Each patient outcome was evaluated with use of multivariate models examining (1) outcomes and hospital-volume strata, and (2) outcomes and surgeon-volume strata. Hospital-volume models were adjusted for surgeon volume as a continuous variable. Surgeon-volume models were adjusted for hospital volume as a continuous variable. All models were adjusted for age, gender, comorbidity index, race, mean income according to the patient zip code, and type of surgical procedure (primary or revision). Point estimates were measured as adjusted odds ratios. We attempted to perform stratified analyses according to the type of surgical procedure (primary or revision), but the numbers of events in the subsets were too small to support multivariate analyses. Instead, we used the type of surgical procedure as a covariate in all multivariate analyses to adjust for this possible confounder. Similarly, the number of important postoperative medical complications, such as myocardial infarction or stroke, was too small to support multivariable analysis and thus could not be assessed.

• We examined bivariate relations between the type of practitioner and each outcome variable, using a one-way analysis of variance or the Kruskal–Wallis test for continuous data, Pearson's chi-square test for categorical data, and nonparametric Kaplan–Meier methods for data on the time to functional recovery. These bivariate analyses were followed by multivariate analyses. A Cox proportional-hazards model was used to estimate the time to functional recovery, with adjustment for covariates that might confound the relation between the type of practitioner and functional recovery. Logistic regression was used for analyses with dichotomous variables. Probabilities and 95 percent confidence intervals were calculated on the basis of model-estimated beta coefficients and standard errors. Multiple linear regression was used to estimate adjusted mean differences in continuous variables among the six strata. Because the cost data were skewed, we modeled charges in three ways: by studying the actual dollar amounts, the log-transformed values, and the rank order of charges. Beta coefficients and standard errors were used to calculate adjusted means and 95 percent confidence intervals.For all analyses, we corrected the standard error for any intragroup correlation due to the cluster sampling scheme(Carey, 1995).

• Our statistical analyses included the χ2 test, Fisher exact test, Student t test, Kaplan–Meier method and log-rank test. We excluded patients who died in the early postoperative period from our survival analysis. We used the Kaplan– Meier method for our survival analyses, and we evaluated the differences using the log-rank test. We evaluated long-term survival and the independent prognostic factors that affect survival using univariate analyses. We performed simultaneous association of multiple variables using the Cox proportional hazards regression model to estimate the simultaneous effect on overall survival. We entered independent variables that showed statistical significance in the univariate analysis into our multivariate analysis. We considered p < 0.05 to be statistically significant. We used the SPSS 10.0 for Windows package (Microsoft) to perform our statistical analyses (Unalp et al,2009)

Missing values

Describe the treatment of missing values and imputation methods, if used.

Statistical software

Describe the statistical software used for all analyses

• All statistical analyses were performed with use of Intercooled Stata (version 6.0; Stata, College Station, Texas) and GNU-R.

• Standard statistical software packages (SAS and Stata) were used for the analyses.(Carey, 1995).

Results

Results from inclusion and exclusion criteria

Differentiate the number of potential eligible patients for the study from the number of those that meet the inclusion criteria. Specify how many cases were excluded due to each exclusion criteria and how many cases were ultimately entered in the study. From these cases, state how many patients stayed in the study until the end, how many withdrew, and provide the reason for withdrawal. To clearly illustrate including a fluxogram is recommended.

Sample description

Describe the population of the study, describing the numbers absolute/percentages, {means/standard error} of the patient socio-demographic characteristics. For example, gender, age, race and other important characteristics to define the sample in the context of the analyzed illness. Show this characterization in a detailed chart. Other clinical variables related to the patients (for example, treatment, description of injuries) should also be described in other tables as necessary.

• Patients and HCC Nodules:Between February 1999 and December 2001, 795 patients with HCC were treated at the Department of Gastroenterology, University of Tokyo Hospital. Among them, 636 were treated with percutaneous RF ablation (clinical profiles shown in Table 1), 224 for primary HCC and the remaining 412 for recurrent HCC. The number of nodules exceeded three in 120 patients (19%), and the nodule diameter was larger than 50 mm in 37 patients (6%). Curative RF ablation was intended in 597 (93.9%) of 636 patients,and a total of 1243 nodules were ablated. In the remaining 39 patients, RF ablation was intended for a total of 176 nodules to reduce tumor burden, with some nodules being left unablated in each patient because of their multiplicity. Thus, a total of 1,419 nodules in 636 patients were treated with ablation. Among them, 231 nodules were in a high-risk location as defined above (Table 3), and a total of 207 patients (32.5%) had at least one nodule in a high-risk location.(Teratani,2006).

• The patients were mostly white (70.2%) and female (62.7%),with a mean age of 68.9 years. The patients treated with revision total knee arthroplasty had a greater burden of comorbid disease, as reflected by a mean Deyo score of 1.7 (95% confidence interval, 1.6 to 1.9) compared with 1.5 (95% confidence interval, 1.4 to 1.6) for the patients treated with primary total knee arthroplasty (Table I)(Hervey, 2006).

• From June 1992 to March 1993, a total of 1633 patients were enrolled in the study. The clinical and demographic characteristics of the patients in the six strata were generally similar, although there were statistically significant differences for a number of variables (Table 1). The patients were relatively young, and a substantial minority had sciatica. Workers' compensation was involved in 31 percent of the cases. In each stratum, 59 percent or more of the patients had acute back pain of less than two weeks' duration. Overall, patients had rapid improvement, with a median of 8 days and a mean of 16 days to functional recovery (a return to a functional status similar to that before the onset of low back pain). Only 5 percent of the cohort had not reported functional recovery by the end of the six-month study period. (Carey, 1995).

• From January 23, 1992 to December 31, 2003, 362 lung transplantations (retransplantations not included) were performed at this center: 228 SLT, 112 DLT (68 bloc-DLT and 44 sequential-DLT), 21 HLT, and 1 lobar lung transplant (Table 1 and Figure 1). Burton CM,2005.

• The average recipient age was 54.0 ± 6.7 years (SLT, 54.8 ± 6.3 years; BSLT, 50.5 ± 7.1 years; p < 0.05), and the average donor age was 30.4 ± 12.7 years (SLT, 30.3 ± 12.8 years; BSLT, 30.5 ± 12.2 years; p = not significant). The average donor ischemic time was 243.2 ± 91.3 minutes (SLT, 222.7 ± 75.5 minutes; BSLT, 329.6 ± 101.6 minutes; p < 0.05). Table I shows patient characteristics for the SLT and BSLT patient groups. The primary diagnosis of the recipient group was emphysema/COPD in all of the patients; patients with α1-antitrypsin deficiency were not included in the study. Meyer, 2001

• Preoperative clinical characteristics: Clinical indices were compared between the two groups (BLT versus SLT) and are shown in Table I. Mean age was lower in the BLT group and sex distribution was significantly different between the groups. Preoperative functional status and comorbidity were evaluated and compared between BLT and SLT. Table I demonstrates that the SLT cohort was functionally different only with respect to preoperative forced vital capacity. The remainder of the pulmonary function test data show no significant difference between groups. Fig. 1, which shows the distribution of the indications for transplantation in the two groups, demonstrates no difference between BLT and SLT with respect to the underlying diagnosis (p = 0.09, not significant). Bavaria, 1997.

Bivariate analysis

Here you describe the results of the statistical tests comparing outcome variables and the main predictors. For example: 1). results of comparing types of chemotherapy for a tumor versus outcomes (reduction of tumor or not), 2). result of comparing the variables number of surgical procedures/year and "hours of hopitalization". This does not take into account the potential confounders so the results are not checked for certain biases. You should always include tables with the results that you have found and preferably mention the p value, along with the confidence interval, so that the reader can quickly identify the statistical significance of the findings.

• Early Complications: The incidence rate of early complications per patient was calculated (Table 4). There were 12 cases (5.8%) of early complications among 207 patients with at least one nodule in high-risk locations, while there were 15 (3.5%) among 429 patients without, the difference was not significant (P=0.1776)(Teratani, 2006).

• Surgeons with combined total knee arthroplasty volumes of sixty procedures or more per year discharged their patients, on average, approximately seventeen hours earlier than surgeons performing fewer than fifteen procedures per year. Patients treated at hospitals with combined total knee arthroplasty volumes of 250 procedures or more per year were discharged approximately fourteen hours earlier than those treated at hospitals with combined total knee arthroplasty volumes of less than eighty-five procedures per year (Table VI). The average length of stay decreased an average of six hours with higher surgeon volumes and an average of five hours with higher hospital volumes, but this relationship was not linear. These values did not display a strictly linear significant trend(Hervey, 2006).

• The median lengths of hospital stay for patients undergoing BCT and BAT were 2.0 days (interquartile range: 1.0 day) and 3.0 days (interquartile range: 2 days), respectively. Postoperative complications were uncommon, affecting 0.5% of BCT and 0.7% of BAT patients. Nonroutine disposition was recorded for 9.7% of patients undergoing BCT and 13.1% of those having BAT. In-hospital mortality was 0.1% for both BCT and BAT (Table 2).

• For all outcomes except in-hospital morbidity after BAT, low-volume providers had the worst outcomes, followed by those with intermediate caseload, while high-volume providers had the best outcomes (Tables 3 and 4).

• Subjects in both PRT and AE groups showed signifififi cant (P<0.05) decrease in plasma glycosylated haemoglobin levels. In terms of percentage, PRT group showed 17.7 per cent decrease and AE group showed 17.9 per cent reduction in HbA1c levels. HDL levels did not change between groups. Triglycerides level decreased by 22.2 per cent in PRT group as compared to 6.9 per cent in AE group and 7.5 per cent decrease in control group. Total cholesterol level significantly decreased in PRT group and in AE group (P<0.05), (Table II). BMI did not change significantly in all three groups after 8 wk. Systolic blood pressure (SBP) was significantly lowered in PRT and in AE groups by 6.5 and 6.2 per cent respectively as compared with control subjects (P<0.05). There was no significant change in diastolic blood pressure (DBP) and heart rate. General well being was assessed by well being questionnaire with a maximum score of 66 and showed a 8.6 per cent increase in score in the PRT group (P<0.05) compared to 2.7 and 2 per cent increase in score in AE and control groups respectively (Table II). (Ekta Arora, 2009).

• Comparison of early postoperative course: Table II summarizes the parameters relevant to the early postoperative course. There was no difference between the SLT and BLT groups with respect to 90-day mortality (SLT 10% vs BLT 7.2%; p = 0.74). Duration of mechanical ventilation (BLT 3.9 ± 1.0 days vs SLT 2.2 ± 0.3 days) and intensive care unit stay (BLT 5.0 ± 1.0 days vs SLT 3.0 ± 0.3 days) were both longer in the BLT group, although neither of these differences achieved significance (p = 0.108 and 0.059, respectively). Duration of hospital stay was similar between groups (BLT 24.5 ± 1.5 days vs SLT 23.4 ± 1.8 days; p = 0.652).Sundaresan RS, 1996.

• Morbidity analysis:We observed no significant differences in the probability of hospitalization for rejection, development of bronchiolitis obliterans, development of bronchial strictures or airway complications, or hospitalization for infection within the first 3 years post-transplant based on procedure type (SLT vs BSLT) and/or recipient age (Figure 5).Meyer, 2001

• Morbidity: Mean length of intubation (BLT 2.43 ± 1.75 days versus SLT 8.13 ± 17.6 days; p = 0.016) was significantly lower in the BLT group, despite the fact that seven recent SLT recipients were extubated on the operating room table. Mean length of in-hospital stay (BLT 35.6 ± 40.4 days versus SLT 30.7 ± 39.5 days; p = 0.31, not significant) did not differ between the two groups. The incidence of major complications was examined and is shown in Table II. No significant difference was detected between the groups except in the case of primary graft failure, which was significantly lower in the BLT group (p = 0.049). Primary graft failure was defined as a form of acute lung injury characterized by the presence of widespread pulmonary infiltrates within lung allograft(s) and associated hypoxemia and ventilatory dependence extending beyond the initial 5 postoperative days. The diagnosis rests on exclusion of other causes of acute lung injury, including pneumonia, aspiration, volumetric overload, hyperacute rejection, massive blood product transfusion, or pulmonary venous outflow obstruction.There was an increased risk of cerebrovascular accident in the BLT group (p = 0.057). Cerebrovascular accident/stroke was defined as any postoperative cerebrovascular accident, including those that resolved without permanent residual deficit. Major gastrointestinal complications were defined as any complication necessitating a laparotomy or therapeutic endoscopic procedure or a gastrointestinal hemorrhage of more than 6 units treated conservatively. Rates of obliterative bronchiolitis, tracheostomy, and bronchial anastomotic complications were no different between procedures. The diagnosis of obliterative bronchiolitis was established on the basis of pulmonary function criteria as recently outlined by Cooper and associates as a decrement in forced expiratory volume in 1 second of more than 20% of posttransplantation baseline.In one case in our series, the presence of obliterative bronchiolitis was confirmed histologically by postmortem investigation of the lung. Additionally, phrenic nerve injury was studied and did not differ between groups. Mortality:The 60-day perioperative mortality rate (BLT 1/29 = 3.45% versus SLT 10/47 = 21.3%; p = 0.03) was lower in the BLT group. Cumulative survivals by Kaplan-Meier analysis, comparing the BLT versus SLT survival curves, were significantly different favoring BLT (p = 0.047) (Fig. 6).Bavaria, 1997

Models

In this section you should present the results describing the association between predictors and outcomes adjusted for potential confounders.

• In risk-adjusted multiple linear regression analysis, length of hospital stay was significantly longer for low-volume hospitals after BCT and BAT compared with high-volume providers (all p<0.001, Table 5).

• Patients who had BCT in low-volume hospitals had an odds ratio of in-hospital mortality of 3.04 (95% CI: [1.12, 8.24], p = 0.03) when compared with high-volume hospitals after risk-adjusting for putative confounding factors (Table 6). For patients who had BAT, the risk-adjusted odds ratios for in-hospital mortality were 1.90 (95% CI: [0.97, 3.70], p = 0.06) for low-volume and 1.78 (95% CI [0.95, 3.30] p = 0.07) for intermediate-volume hospitals compared with high-volume hospitals (Table 6). Low-volume hospitals had significantly higher postoperative complications for both BCT (OR = 1.73, 95% C1: [1.17, 2.56], p = 0.01) and BAT (OR = 1.44, 95% CI: [1.21, 1.72], p < 0.001 when compared with high-volume hospitals.

• Logistic regression was used to determine the effect of the type of provider on the probability of complete recovery at each interval, with an adjustment for clinical and demographic covariates. Thirty-one percent of the patients did not consider themselves completely better at the end of six months. Differences in the probability of recovery according to the type of provider were of borderline significance at base line and 2 weeks and not significant at 4, 8, 12, and 24 weeks. After adjustment for base-line variables, there were no significant differences among the six strata in the estimated mean disability scores on the Roland–Morris scale. Patients who reported functional impairment continued to have high disability scores.Over 95 percent of the patients were back at work by the four-week interview, with no differences among the strata.The data presented are unadjusted; adjustment for the base-line variables did not substantially affect the results.The mean number of visits differed substantially among the strata (Table 2), ranging from 3.1 in the HMO stratum to 4.4 and 5.5 in the primary care and orthopedic strata, respectively. The mean number of visits among the patients seeing chiropractors was 10.1 in the rural stratum and 15.0 in the urban stratum. Although the most common treatment used by chiropractors was spinal manipulation, additional treatments included heat, cold, diathermy, ultrasonography, electrical stimulation, and traction.The average number of prescription or over-the-counter medications used was higher among the patients seen by primary care physicians, orthopedists, or HMO providers than among the patients seen by chiropractors (3.5 vs. 2.3 medications, P<0.001).Plain spine radiographs were used more frequently by the chiropractors and orthopedists (in 67 to 72 percent of the patients) than by the other groups of providers. There were no statistically significant differences in the use of radiographs among the three primary care strata.We estimated the charges for outpatient care in each of the six strata (Table 3). Both unadjusted and adjusted data are presented.The distribution of charges per episode of back pain was highly skewed, with a small proportion of patients (6 percent) who had outpatient charges exceeding $2,000. The median charge per episode, which is less subject to the effect of these outliers than the mean charge, shows the same pattern, with higher median charges in the strata of orthopedists and chiropractors (Table 3). Charges in the three primary care strata were very similar. The HMO stratum had fewer outliers than the other strata (3 percent of the HMO patients had charges exceeding $2,000). Log transformation and nonparametric tests yielded a similar pattern.atients who saw chiropractors reported a significantly higher degree of satisfaction than those who saw practitioners in the other four strata (Table 4). A logistic-regression analysis showed that the patients who saw orthopedic surgeons were somewhat more satisfied than the patients who saw primary care providers but were less satisfied than those who saw chiropractors. The higher level of satisfaction among the patients who saw chiropractors persisted after adjustment for the number of visits and the use of radiography. The strongest correlates of satisfaction were the patient's responses to questions about the quality of the provider's history taking, examination, and explanation of the problem during the visit (Table 4). (Carey, 1995).