There has been significant pressure to lower the cost of health care delivery in the United States for the last 20 years. Various payers from Medicare to private insurers have identified hospital costs as 1 of the major components of the overall cost of care. They have introduced multiple incentives over time, which have been mostly based on changes in reimbursement policies, to pressure hospitals into more efficient practices.1 One of the constant goals of hospitals and health systems has, then, been to reduce the duration of patients’ hospitalization. Although significant progress have been made in lowering the length of stay in most of the high-income countries, the length of stay for acute care in the United States has stabilized at 4.9 days without any change between 2000 and 2009.2 Multiple specific interventions have been proposed to further decrease the length of hospital stay. However, few have addressed 2 common barriers: fragmentation of care and unwillingness of physicians and health care providers in general to implement measures that they did not design themselves.3
The Perioperative Surgical Home (PSH) has been recently proposed as a model to improve the delivery of health care via patient-centered optimization strategies that involve risk stratification and standardization of care. The PSH model hopes to reduce variability in perioperative care given that these inconsistencies increase the likelihood for errors and complications.4
Although the PSH model has been implemented at a few adult hospitals, there are no data on its implementation in the pediatric surgical population.5
Children with scoliosis requiring a posterior spinal fusion (PSF) are an ideal population to test the potential benefits of applying the principles of the PSH. Adolescent idiopathic scoliosis is the most common spinal condition requiring surgical treatment during teenage years, and the pre-/intra-/posthospital care of these patients requires the intervention of multiple subspecialties. Moreover, despite a significant decrease in the last decade, the hospital length of stay after PSF still averages 5.6 days in the United States.6
The aim of this review was to determine whether the implementation of a new care protocol for children requiring a PSF, based on the principles of the PSH and designed by every stakeholder involved in these patients’ care, could lower the hospital length of stay and hospital costs without affecting the quality of care through improved care coordination and preoperative patient education.
In June 2014, a team of anesthesiologists, orthopedic surgeons, nurses, and physical therapists analyzed every aspect of the care delivered to patients undergoing a PSF at the Children’s Hospital Los Angeles during the previous 2 calendar years. Based on their findings, they designed a new protocol with the goal of lowering the hospital length of stay and hospitalization costs when maintaining the quality of care and patients’ satisfaction that was implemented in January 2015.
Patients and families were given a tour of the hospital facility 2 to 3 weeks before their admission. On the same day, nurse practitioners from the orthopedic and pain services provided information on the patient’s expected perioperative course and services that they would receive including: details on the surgical procedure, medications, dietary, and physical therapy interventions (Table 1). Explanations were given to patients and their families with respect to what the patient could experience in terms of pain severity and pain quality (incisional, muscle spasm, etc) as well as the pain control methodologies used to minimize them.
On the day of surgery, the anesthesia team would first meet the patient and obtain informed consent for anesthesia.
The intraoperative anesthetic management was standardized and included: induction with 2 mg IV midazolam followed by 1 mg/kg ketamine up to 50 mg and 2 mg/kg propofol. Muscle relaxation was achieved using 0.05 mg/kg cisatracurium followed by endotracheal intubation. Tranexamic acid was given as an initial bolus of 50 mg/kg over 30 minutes and then continued as an infusion of 5 mg/kg/h for the duration of the surgery. Somatosensory-evoked potentials and motor-evoked potentials were monitored in every patient. Isovolemic hemodilution was done if deemed appropriate by the attending anesthesiologist. Fentanyl at 1–2 µg/kg was given before incision and preservative-free morphine given intrathecally by the surgeon at a dose of 5 µg/kg up to 250 µg. Maintenance was via total IV anesthesia using starting doses of 250 µg/kg/min propofol and 0.2 µg/kg/min remifentanil titrated to maintain Bispectral Index of 55 to 60. We varied the target blood mean arterial pressure (MAP) according to the phase of surgery: phase 1 (cutdown/dissection and decortication of vertebral lamina): MAPs 55 to 60 mm Hg, phase 2 (placement of pedicle screws): MAPs 60 to 65 mm Hg, and phase 3 (spinal distraction): MAPs 75–80 mm Hg. The propofol and remifentanil were turned off after final neuromonitor testing and at the closing of skin, respectively. Acetaminophen IV (15 mg/kg) was given at the end of the procedure. The antiemetic protocol included the intraoperative administration of ondansetron (0.1 mg/kg, maximum 4 mg) and dexamethasone (0.15 mg/kg, maximum 8 mg). After the demonstration of movement of bilateral lower extremities to the satisfaction of the orthopedic surgery provider, the patient was extubated and transported to the recovery room.
The surgical technique was also standardized with the ultimate goal to decrease the intraoperative surgical time. Verification of pedicle screw placement via fluoroscopy was limited to 2 (first and last), and the pedicle screws were placed and the screw heads were removed using a power tool (POWEREASE System; Medtronic, Minneapolis, MN) rather than done manually.
The pain management protocol included the use of morphine or hydromorphone delivered via patient-controlled analgesia (PCA). The initial doses were standardized. Morphine demand doses were 0.20 mg/kg every 6 minutes; hydromorphone doses were 0.04 mg/kg; and the continuous infusion was set at 15 µg/kg/h for morphine and 3 µg/kg/h for hydromorphone. The doses were then titrated on the basis of patients’ pain level and side effects. The choice between the 2 opioids was based on patients’ previous experience with each opioid and/or their response to the initial treatment in the immediate postoperative period. IV diazepam (0.1 mg/kg, every 6 hours, maximum 5 mg) was started in the recovery room and ketorolac (0.5 mg/kg, maximum 30 mg, every 6 hours) was started on postoperative day 0 unless patients had >300 mL of intraoperative blood loss. The PCA was stopped, and patients were transitioned to oxycodone (0.1 mg/kg, maximum 10 mg) in the afternoon of postoperative day 1 unless they were experiencing repeated episodes of vomiting (>2 in the last 24 hours) or severe nausea such as they would refuse to take solid food. Similarly, the diazepam was switched from IV to an oral formulation in equivalent dose for patients requiring liquid formulations or rounded to either 2 or 4 mg for patients requesting tablets.
Patients were encouraged to start a liquid diet on postoperative day 0 and were then given a high-fiber diet on postoperative day 1. Patients received IV ondansetron (0.1 mg/kg, maximum 4 mg, every 8 hours) for the initial 24 hours after surgery. The same dose of ondansetron was then administered orally as needed for the rest of the hospital stay.
A bowel regimen was started in all patients on postoperative day 0 and included docusate (100 mg twice daily), bisacodyl (10 mg twice daily), and Miralax (Bayer, Whippany, NJ) (17 g daily). Physical therapy was started on postoperative day 0 with the patients being asked to sit at the edge of the bed. Patients were asked to walk 2 to 3 times on postoperative day 1, initially inside the room and then in the hallway, with the assistance of the physical therapist and unit nurses. Patients were then asked to walk longer distances with a goal of 500 ft on postoperative day 2 and to climb stairs (goal of 1 flight), if tolerated.
Teams of anesthesiologists, orthopedic surgeons, pain specialists, nurse practitioners, and floor nurses were created and educated about the protocol details. The care of patients who underwent a PSF after the creation of the new protocol was then assigned to these trained teams.
We reviewed the electronic medical record of patients who underwent a PSF for idiopathic scoliosis from January 2013 to September 2015. On July 1, 2014, we implemented the new treatment protocol for patients undergoing a PSF. We then identified 2 groups of patients: those who underwent a PSF between January 2013 and June 2014 and those who underwent a PSF from July 2014 to September 2015. We compared demographic and outcomes data before and after July 2014, the time of the implementation of the new patients’ care protocol. Patients were excluded from the analysis if they had surgical complications such as loss of neuromonitoring signals, dural punctures or tears, or others requiring admission to the postoperative intensive care unit. Patients who required prolonged hospitalization because of concomitant medical conditions including the presence of neoplastic lesions, mental retardation, cardiac malformations, previous spine operations, spondylolisthesis, and those who had a <6-level fusion were also excluded from the analysis.
Demographic data included age, sex, Cobb angle, and American Society of Anesthesiologists status. Intraoperative data included the use of intrathecal morphine, IV ketamine use, estimated blood loss, and surgical time. The following postoperative data included postoperative morphine equivalents used, PCA duration, time to first solid food, the presence of vomiting within the first 24 hours, total length of stay, time to ambulation, and bowel movement. Pain levels were assessed using a 0 to 10 Verbal Numeric Scale.7 The total dose of opioids, IV and oral, used during the hospitalization is reported as daily dose of morphine equivalent opioids. The conversion factors are shown in Table 2.8
Duration of surgery was calculated on the basis of anesthesia records from the time of incision to the time patient was extubated. Hospital stay was calculated from the time the patient left the operating suite to the time at which the nursing staff formally discharged the patient from the hospital floor. Complications were determined based on surgical operative notes.
Discharge criteria included tolerating oral medications and no need for IV medications, log rolling and lying to sitting, sitting to standing and pivoting, ambulating 500 ft and going up and down 1 flight of stairs, pain being controlled by oral medications, and positive flatus. Having a bowel movement was not required before being discharged home. The decision to discharge the patients home was made by the entire care team, surgeons, physical therapists, and pain specialists.
Patient follow-up was conducted by the orthopedic surgeons. Data that were retrieved from their notes included duration of opioid intake, time to return to preoperative level of physical activity, including sports, and surgical complications. Every patient was seen 2 weeks after the hospital discharge and the following visits were scheduled on the basis of the patients’ clinical needs.
Normality of each variable was assessed using the Shapiro–Wilk test of normality and was also based on the skewness and kurtosis of the residuals distribution. Differences between continuous variables were analyzed using the 2-sample t test or a Mann-Whitney U test based on their distribution. Categorical variables were analyzed using χ2 analysis. The Fisher exact test was used when required. Šidák-Holm–adjusted P values have been reported when indicated.
A multivariable least square regression model was used to analyze the independent relationship of our main effect (new protocol) on the overall total length of stay. In particular, we looked at potential predictors associated to a length of stay of 3 days or less. Potential predictors for inclusion in our model were selected on the basis of their independent relationship to the new or old protocol such as the duration of surgery, number of levels fused, surgeon, age, sex, estimated blood loss, and Cobb’s angle. Potential predictors were dropped from the model in a backward elimination approach at each level by assessing whether each predictor’s statistical significance was lacking (using a cutoff of P < .1). Colinearity among predictors in our main effect model was assessed by its variable inflation factor and removed as appropriate. The selection of the optimal model at each step was assessed by Akaike information criterion and Schwarz Bayesian information criterion.
The difference in length of stay over time between the 2 protocols was calculated using the Mann-Whitney U test. We also compared the 2 protocols on the proportion of patients who stayed more than 3 days in the hospital using a χ2 test.
We compared the 2 protocols on the change over time in oral and IV opioids by fitting separate regression lines for each period and constructing the Z test based on slope and standard error estimates.
A priori power analysis was conducted to determine the power of the new intervention to have a statistically significant effect on the variables we analyzed.
Assuming a standard deviation in length of stay of 1.5 days based on a sample of patients who had a PSF in the last trimester of 2012, it was estimated that a sample size of 36 patients per group would be needed to detect a 1-day difference in length of stay at 80% power and α level of .05.
The study was approved by the Children’s Hospital Los Angeles institutional review board. One hundred eighty-eight patients underwent a PSF during the study interval. Patients were chronologically selected starting from the study time 0 until 36 patients per group who met the inclusion criteria were identified. We analyzed the data of 40 patients managed before the implementation of the protocol and 40 patients managed with the new protocol. Four patients in each group were excluded from the final analysis because they did not meet the inclusion criteria. Demographic data were similar between the 2 groups with the exception of a higher percentage of patients with an American Society of Anesthesiologists status III in patients managed under the new protocol guidelines (P = .003; Table 3).
Patients were operated on by 4 different surgeons (Table 4). Surgeon 4 operated on 64% and 61% of the patients in the old and new protocol, respectively. The distribution of cases among the 4 surgeons was similar between the 2 groups of patients. Similarly, there were no differences in length of stay among the 4 surgeons (old protocol: P = .07; new protocol: P = .16). A similar number of patients received intrathecal Duramorph and IV ketamine during the operation, and estimated blood loss was also similar between the 2 groups (Table 4). The number of levels fuses was significantly higher in the new protocol group (P = .017; Table 4). A significantly greater number of patients underwent isovolemic hemodilution in the new protocol group (P = .001). The duration of surgery was similar in the old versus new protocol groups (P = .21). Surgery was conducted before 11:00 am in 32 patients from the old protocol group and in 30 patients in the new protocol group (P = .2). Procedures were never started after 2:00 pm.
PCA use was significantly decreased in the new protocol group (24 ± 8 vs 58 ± 22 hours; P= .001). PCA was stopped on postoperative day 2 in 6 patients (17%) because of nausea or vomiting. The cumulative dose of morphine equivalent used by patients was significantly lower in the new protocol group on postoperative days 2 and 3 (Table 5). There was no significant difference in the use of oral and IV opioids between the 2 study groups (Figure 1 A, B). The time to first solid food intake (24 ± 8 vs 39 ± 18 hours; P = .001) was significantly shorter in the new protocol patients; 64% were eating solid food within 24 hours from the operation versus 22% of the patients in the old protocol group (Table 5). Similarly, the time to first ambulation with physical therapy (23 ± 7 vs 36 ± 18 hours) was significantly shorter in the new protocol group; 86% were ambulating within 24 hours from the operation versus 42% of the patients in the old protocol group (Tables 4 and 5).
The hospital length of stay was 3.9 ± 0.7 days (range 3–5 days) for patients treated with the old protocol and 3.3 ± 0.7 days (range 2–4 days) in the group treated with the new protocol (P= .001; Table 5). The regression analysis of the 2 slopes in Figure 2 showed a significant steeper line in the new protocol group (P= .001). When breaking down the time of discharge from the hospital, 3 patients (8%) managed with the new protocol were discharged home on postoperative day 2, whereas none of the patients managed with the older protocol met the discharge criteria by the postoperative day 2 (Figure 2). Seventeen patients (47%) in the new protocol group were discharged home on postoperative day 3 vs 11 patients (30%) who had been managed with the old protocol (Figure 2). No patient in the new protocol group was discharged on postoperative day 5 vs 20% of the patients in the old protocol group (Figure 2). A significantly greater percentage of patients was hospitalized for more than 3 days in the old versus the new protocol group (69% vs 44%; P = .032) (Figure 3).
In our adjusted multiple regression model, patients who were managed with the new protocol had a 0.648-day (95% confidence interval [CI], −0.982 to −0.370) reduction in overall total length of stay (P = .001). Patients who had 10 to 12 levels or greater than 12 levels of fusion had a 0.396-day (95% CI, 0.009–0.782) and 0.448-day (95% CI, 0.045–0.850) longer total length of stay, respectively, compared to those who had less than 10 levels of fusion (Table 6).
There was no statistical difference between the 2 groups when looking at the number of patients who reported pain scores >3 or severe pain (scores >5) throughout the hospitalization (Table 7).
Follow-up visits were conducted in the orthopedic surgery outpatient clinic. Data were available for every patient considered in the study. As expected, the duration of the follow-up was of longer in the old protocol group (56 ± 48 months) compared to the new protocol group (18 ± 9 months; P = .0001). There was no difference between the 2 groups with respect to the number of weeks patients took oral opioids during the follow-up period (old protocol 2.5 ± 1.9 vs 2.7 ± 2 new protocol; P = .8). The return to physical activities, including sports, was significantly shorter in the new (5.9 ± 3.2) versus the old protocol group (8.2 ± 7.2; P = .0001). One patient in the old protocol group was sent home with a wound drain, and 1 developed a wound infection. Four patients were sent home with a wound drain, and 1 patient developed a wound infection during the follow-up period. Infectious complications were managed at home with oral antibiotics and dressing changes. No patient was readmitted to the hospital during the follow-up period.
Data from this review show that a coordination of care among the different stakeholders involved in the care of children undergoing a PSF for scoliosis can significantly reduce the length of hospital stay and its related costs. The creation of protocols directing the care of these patients in the preadmission phase, during and after the hospital stay, seems to be critical in assuring a comfortable and uncomplicated transition between the different phases of this complex surgical procedure.
PSH is modeled to improve delivery of quality, coordinated care in a cost-effective manner to patients undergoing surgical procedures.9 Improving these aspects of health care is becoming increasingly important in the current climate of bundled payments and patient satisfaction surveys. As a result, fast-track or accelerated discharge pathways have been implemented in adult surgical patient populations such as in joint replacement programs.10–12 The goal of creating a new care protocol in children undergoing a PSF was to lower the length of stay without compromising other aspects of their care.
It is important to recognize that the role of the PSH cannot and is not limited to simply reducing the overall costs of care. It should also take into consideration the experience that patients and their families live through once a decision to proceed with an operation is made. There is a host of indicators that could be used to measure quality of care. As the Centers for Medicare & Medicaid Services is still trying to identify ideal indicators, we selected postoperative pain because this is a major concern of patients undergoing an operation.13 Our data show that it is possible to reduce the length of stay without compromising pain experienced by patients and the rate of readmission.
The results of our study are in line with Fletcher et al14 and Shan et al,15 who have shown the safety and economic impact of an accelerated discharge model in idiopathic scoliosis posterior spinal fusion surgeries. However, no measurements of pain scores were provided in those studies.14,15 Alternatively, Muhly et al16 have shown that multimodal analgesia in the setting of an accelerated discharge model also reduced length of stay to 4 days average without adversely affecting pain scores.
There was no statistically significant difference in the number of patients reporting moderate and severe pain between our 2 treatment protocols. In particular, on postoperative days 3 and 4, although not statistically significant (our data had only a 44% power to detect a significant difference), more patients in the new protocol group experienced moderate pain compared to those managed with the old protocol. This difference may be related to the early mobilization of the patient under new protocol guidelines.
The length of stay was found to be significantly decreased after the initiation of the new protocol. We reported the number of patients discharged on each postoperative day in addition to the average length of stay because we felt that provided more informative data regarding the distribution. The average length of stay in the old protocol versus the new protocol is in line with published accelerated discharge models.14,17 However, when looking at overall saving costs, it is more important to look at the actual number of days patients spend in the hospital because of the way hospitals are reimbursed for their services. The management of the same number of patients with similar demographics utilizing the new protocol resulted in saving 19 days of hospital costs charges, as reported by hospital business administration.
The entire care team previously described participated in the design of the new protocol. It is conceivable that they felt pressure to demonstrate the success of the new approach when managing patients with the new protocol. However, the discharge criteria, listed in the “Methods” section, did not change between the 2 protocols, and they are strictly based on patients’ performance and not influenced by care providers.
The key to the success of the implementation of the new protocol to manage children who require PSF depended on our multidisciplinary approach to caring for these patients. The statistical analysis suggests that more than any 1 single intervention, the implementation of the new multifaceted protocol was associated with a shorter length of stay. The most important change introduced with the protocol was the standardization of the postoperative aspect of care, which included pain management, diet, and physical and occupational therapy protocols. Patients’ and families’ expectations of treatment may also be an important determinant of the observed outcomes. It is theorized that the preoperative setting of expectations and continued encouragement of ambulation during the hospital stay leads to continued improvement postoperatively.18 It is conceivable that some of the results could be attributed to the preoperative information sessions, which had the goal of providing a better understanding of expectations and the overall processes associated with a PSF hospitalization. Although we did not measure the impact of this specific change on the overall results, it would be important to confirm the value of patient and family education in future studies.
The long-term follow-up data showed that pain was not a major problem with patients stopping oral opioids within 2.5 weeks after the operation. It is unclear why the return to what children considered normal physical activities was significantly shorter in the new protocol group.
This study has several limitations. It was an interrupted time-series design without data from a control population. We analyzed data from a relative small number of patients. Despite the limited number of providers involved in these patients’ care over time, this was a retrospective study with the possibility for biased or inaccurate data collection. Although it could be argued that the results were influenced by the care teams that designed the new protocol, the discharge criteria, listed in the “Methods” section, did not change between the 2 protocols; they were objective criteria strictly dependent of patients’ performance and not influenced by the care providers.
Although we did not have a study specific patient satisfaction survey, pain scores on the day of discharge were used as a surrogate for patient satisfaction with regard to pain during the hospital stay. Specific patient satisfaction surveys with longer follow-up may reveal areas of further improvement. Further studies should include longer follow-up and more detailed information on the postoperative quality of life of these patients as well as potential interventions that may help these children return to their preoperative activities in a shorter period of time. In addition, pain scores during the postoperative period were still elevated in approximately 25% of the patients. The routine use in the future of additional adjuvant medications such as gabapentin or clonidine to better control pain may also help accelerate the patient’s discharge process while improving their pain.
In conclusion, we assessed our posterior spinal fusion protocol in the context of a surgical home model. We were able to significantly decrease length of stay and hospital cost while not adversely affecting pain scores. In addition, with re-evaluation of the protocol, we were able to identify areas of improvement. We have implemented these changes and will continue to re-evaluate and improve patient care using a Surgical Home model.
Name: Eugene Kim, MD.
Contribution: This author helped conceive and design the study, acquire, analyze and interpret the data, and draft the manuscript.
Name: Brian Lee, MD.
Contribution: This author helped conceive and design the study, acquire, analyze and interpret the data, and draft the manuscript.
Name: Giovanni Cucchiaro, MD.
Contribution: This author helped conceive and design the study, analyze and interpret the data, and revise the manuscript.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
1. Southern WN, Arnsten JH. Increased risk of mortality among patients cared for by physicians with short length-of-stay tendencies. J Gen Intern Med. 2015;30:712–718.
2. Organization for Economic Cooperation and Development. Average Length of Stay in Acute Care; OECD Health Data 2011.
Paris, France: OECD; 2011.
3. Borghans I, Kool RB, Lagoe RJ, Westert GP. Fifty ways to reduce length of stay: an inventory of how hospital staff would reduce the length of stay in their hospital. Health Policy. 2012;104:222–233.
4. Kain ZN, Vakharia S, Garson L, et al. The perioperative surgical home as a future perioperative practice model. Anesth Analg. 2014;118:1126–1130.
5. Correll DJ, Bader AM, Hull MW, Hsu C, Tsen LC, Hepner DL. Value of preoperative clinic visits in identifying issues with potential impact on operating room efficiency. Anesthesiology. 2006;105:1254–1259.
6. Daffner SD, Beimesch CF, Wang JC. Geographic and demographic variability of cost and surgical treatment of idiopathic scoliosis. Spine (Phila Pa 1976). 2010;35:1165–1169.
7. Bailey B, Daoust R, Doyon-Trottier E, Dauphin-Pierre S, Gravel J. Validation and properties of the verbal numeric scale in children with acute pain. Pain. 2010;149:216–221.
8. Shaheen PE, Walsh D, Lasheen W, Davis MP, Lagman RL. Opioid equianalgesic tables: are they all equally dangerous? J Pain Symptom Manage. 2009;38:409–417.
9. Vetter TR, Goeddel LA, Boudreaux AM, Hunt TR, Jones KA, Pittet JF. The Perioperative Surgical Home: how can it make the case so everyone wins? BMC Anesthesiol. 2013;13:6.
10. Raphael M, Jaeger M, van Vlymen J. Easily adoptable total joint arthroplasty program allows discharge home in two days. Can J Anaesth. 2011;58:902–910.
11. Garson L, Schwarzkopf R, Vakharia S, et al. Implementation of a total joint replacement-focused perioperative surgical home: a management case report. Anesth Analg. 2014;118:1081–1089.
12. Jones S, Alnaib M, Kokkinakis M, Wilkinson M, St Clair Gibson A, Kader D. Pre-operative patient education reduces length of stay after knee joint arthroplasty. Ann R Coll Surg Engl. 2011;93:71–75.
13. Royston D, Cox F. Anaesthesia: the patient’s point of view. Lancet. 2003;362:1648–1658.
14. Fletcher ND, Shourbaji N, Mitchell PM, Oswald TS, Devito DP, Bruce RW. Clinical and economic implications of early discharge following posterior spinal fusion for adolescent idiopathic scoliosis. J Child Orthop. 2014;8:257–263.
15. Shan LQ, Skaggs DL, Lee C, Kissinger C, Myung KS. Intensive care unit versus hospital floor: a comparative study of postoperative management of patients with adolescent idiopathic scoliosis. J Bone Joint Surg Am. 2013;95:e40.
16. Muhly WT, Sankar WN, Ryan K, et al. Rapid recovery pathway after spinal fusion for idiopathic scoliosis. Pediatrics. 2016;137:e20151568.
17. Fletcher ND, Andras LM, Lazarus DE, et al. Use of a novel pathway for early discharge was associated with a 48% shorter length of stay after posterior spinal fusion for adolescent idiopathic scoliosis. J Pediatr Orthop. 2017;37:92–97.
18. Thompson AG, Suñol R. Expectations as determinants of patient satisfaction: concepts, theory and evidence. Int J Qual Health Care. 1995;7:127–141.