Introduction 1 2 3 4 5 6 8 9 13 14 15 16 22 15 Therefore, our aims were to determine the presence and level of symptoms of PTSD in patients surviving abdominal sepsis. In addition, we searched for demographic and disease-related factors associated with higher levels of PTSD symptoms. Identification of such factors may be important to determine possible targets of intervention and to select patients for psychological assessment interviews. Methods Study design Our study was embedded in an ongoing randomized clinical trial (the RELAP Trial) evaluating two surgical treatment strategies for patients with secondary peritonitis after the initial emergency laparotomy. Patients were enrolled between December 2001 and February 2005 in two academic medical centers and seven regional teaching hospitals in The Netherlands. All patients were followed up for 12 months after initial (index) laparotomy. The study was approved by the medical ethics committee of the Academic Medical Center, Amsterdam. All patients gave informed consent to participate in this study. Study population 23 Data collection All self-administered PTSD questionnaires were distributed by mail to patients who survived at least 12 months following initial emergency laparotomy, with a reminder by phone within 2 weeks in the case of no response. After 1 month without response a new questionnaire including a reminder letter was sent. Instruments assessing the level of PTSD symptoms 24 25 26 27 28 Post-Traumatic Stress Syndrome Scale 10 14 4 11 12 11 29 The Impact of Events Scale–Revised 27 28 28 30 27 Potential risk factors 31 17 22 32 33 6 9 11 14 34 35 General patient characteristics Disease characteristics and postoperative course 36 13 37 4 12 20 23 Traumatic memories of ICU/hospital stay 13 38 38 Data analysis 39 40 p 34 In addition, a factor comprised of other non-related traumas that the patient had experienced within the previous 3 years was included in the final model to assess its potential confounding role. The fit and validity of the model was evaluated by checking the discriminatory properties (overlap in risk scores of patients with different outcomes), the proportional odds assumption (test for parallel lines) and calibration (closeness in expected and observed numbers of patients evaluated by an extension of the Hosmer–Lemeshow goodness-of-fit statistic). 41 Nomogram: A nomogram was developed to visualize the prognostic strength of the different factors from the multivariate model in a single diagram. A nomogram allows readers to calculate an expected distribution of PTSD symptomatology (‘low-scoring’, ‘moderate-scoring’ and ‘high-scoring’ patients) based on a specific profile of a patient. The number of points for each predictor was based on the original coefficient from the multivariate ordinal model. The total number of points derived by specifying values for all predictors was used to calculate the expected probabilities that a patient would be a ‘low-scoring patient’, a ‘moderate-scoring patient’ or a ‘high-scoring patient’. Analyses were performed using SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA). Results 1 Fig. 1 Flowchart summarizing inclusion and response  1 Table 1 Association between severity of PTSD symptoms (three categories) and patient, disease operative and postoperative characteristics: results from univariate ordinal regression models  Overall a n b   n n n p General patient characteristics Median age (IQR) 66.8 (57–73) 70.2 (60–74) 58.7 (47–72)  57.8 (49–65) 0.004 Male gender (%) 54% 53% 53%  64% 0.847 c 53% 55% 50%  55% 0.670 Peritonitis and postoperative characteristics Initial       Median APS score (IQR)  6 (4–8)  6 (4–8)  7 (5–9)  8 (3–8.5) 0.271  Hydrocortisone in first 14 days in ICU (median days)  2 (0–7)  1.5 (0–8)  1 (0–8)  5 (1–7) 0.749  ARDS  6%  3% 10%  9% 0.192  One or more relaparotomies 67% 70% 63%  64% 0.515  Admitted to ICU 89% 85% 93% 100% 0.110  Median length of ICU stay (IQR)  7 (4–15)  7 (4–12)  7 (4–19)  9 (6–16) 0.042  Median ventilation time (IQR)  5 (1–8)  4 (1–7)  5 (1–10)  7 (4–13) 0.073  Median length of hospital stay (IQR) 28 (19–55) 26 (18–47) 31 (23–60)  56 (19–72) 0.102 Follow-up       Disease-related major morbidity at 6-month follow-up 15%  9% 27%  18% 0.068  Enterostomy at 6-month follow-up 51% 47% 55%  70% 0.183 IQR a b c Prevalence of PTSD symptoms 1 Predictive factors 1 2 Table 2 Association between severity of PTSD symptoms (three categories) and other traumatic experiences following peritonitis  n Univariate ordinal regression  n a n a n p Traumatic memories of ICU or hospital stay Nightmares 39% 61%  82%       0.002 Fear and panic 24% 61% 100% < 0.001 Pain 67% 70%  82%       0.002 Difficulty breathing 33% 76% 100% < 0.001 Traumatic memories None (0) 41% 50%  9% < 0.001 Moderate (1–4)  7% 47%  47%  Severe (> 4)  0% 18%  82%  a 3 Table 3 Association between severity of PTSD symptoms and patient, disease operative and postoperative characteristics and other traumatic experiences following peritonitis in a multivariate analysis  n a  OR 95% CI p   Lower Upper  Ten years increase in age  0.74 0.53  1.04       0.084 Female  0.9 0.94  2.3       0.822 Length of ICU stay (log2 transformed)  1.4 1.1  1.7 < 0.003 Major disease-related morbidity during 6-month follow-up (including index hospital admittance)  2.1 0.61  7.11       0.238 Traumatic memories of ICU or hospital stay Moderate (1–4)  4.9 0.95  24.9       0.058 Severe (> 4) 55.5 9.4 328.0 < 0.001 Other trauma within previous 3 years  2.4 0.94  6.3       0.085 a p p p p p p p 2 Fig. 2 Nomogram for prediction of severity of PTSD symptoms in patients with secondary peritonitis. Graded outcome categories are: none to mild (negative on both instruments), moderate (positive on one instrument), and severe (positive on both instruments)  p p 3 p Fig. 3 Distribution of total points from nomogram (risk score) for the prediction of the severity of PTSD symptoms with use of the risk factors taken from the multivariate ordinal model. PTSD categories are graded according to severity: none to mild (negative on both instruments), moderate (positive on one instrument), severe (positive on both instruments)  Discussion 9 13 15 42 7 43 44 36 15 34 13 34 15 34 12 45 10 12 15 13 37 46 47 48 48 9 49 11 29 50 48 51 50 49 52 9 48 53 54 In the clinical setting, there is a continuing debate on whether to intervene in the more acute peritraumatic psychological processes or in a later phase, when symptoms or prodromes of PTSD are observed. By improving our understanding of which factors play an important role in the development of PTSD, we can better prevent PTSD symptoms in high-risk patients and decide when best to intervene. The aim of our predictive model is for it to be used by treating physicians, following the acute episode and phase of secondary peritonitis in which survival and physical recovery are the main concerns, to recognize high-risk PTSD patients. This relatively simple model can aid the surgeon, for instance, during the first outpatient visit in determining which patients are at higher risk for the development of PTSD symptoms. However, before this nomogram can be used to actually predict PTSD symptomatology in clinical practice, it must be externally validated in another cohort of patients with secondary peritonitis. In conclusion, 10% of peritonitis patients report ‘high’ PTSD symptomatology and another 28% ‘moderate’ PTSD symptoms. Factors that were related to more PTSD symptoms included younger age, traumatic memories of the period of hospitalization and length of ICU stay. Knowledge of these predictive factors is required to increase awareness, and to develop tailored early treatment options for these high-risk patients our nomogram may assist in identifying patients with PTSD symptoms. Electronic supplementary material Electronic Supplementary Material (DOC 21K) Electronic Supplementary Material (DOC 52K)